EP4681397A1

EP4681397A1 - Randomizing deep neural networks for telecommunications

Info

Publication number: EP4681397A1
Application number: EP24740718.2A
Authority: EP
Inventors: Jibing Wang; Erik Richard Stauffer
Original assignee: Google LLC
Current assignee: Google LLC
Priority date: 2023-06-16
Filing date: 2024-06-14
Publication date: 2026-01-21
Also published as: WO2024259268A1; CN121241545A; KR20260003173A

Abstract

End-to-end deep neural network (DNN) communications are provided between a first device (104a) and a second device (104b). At the first device, a transmitting DNN (106) processes input communication data to generate an output communication signal (118). The first device performs a scrambling DNN operations in response to a forecasted transmission of the generated output communication signal not satisfying a white noise interference level, including selecting neural network scrambling information (NNSI) for reconfiguring the transmitting DNN to process the input communication data to generate a scrambled output communication signal, which when transmitted satisfies the white noise interference level. The first device transmits a control message including an indication of the NNSI and scrambling timing information for directing when the second device is to reconfigure a corresponding receiving DNN (109) for generating reconstructed communication data. The first device transmits the scrambled output communication signal based on the scrambling timing information.

Description

RANDOMIZING DEEP NEURAL NETWORKS FOR TELECOMMUNICATIONS

BACKGROUND

[0001] Conventional fourth generation (4G) and fifth generation (5G) communication systems have complicated transmitter and receiver processing chains with multiple processing components for encoding and modulating input communication data for wireless transmission from a first device and reception and reconstruction by a second device. The complexity of current transmitter and receiver processing chains can be reduced by training machine learning (ML) algorithms, such as deep neural networks (DNNs), to form a transmitter (TX) DNN model (TX DNN) and a receiver (RX) DNN model (RX DNN) (also referred to as a transmitting DNN and receiving DNN) capable of providing end-to-end communications. Such transmitter and receiver DNN models potentially augment and/or replace conventional transmitter and receiver processing chains. For example, a trained transmitting DNN generates transmission waveforms suited to efficiently overcome numerous channel environments, impairments, and interference found in current communication systems (e.g., multi-path interference, multiple access interference, narrowband interference) further enhancing performance. In addition, such transmitting and receiving DNN models are well suited for supporting end-to-end communication systems for where it may be impractical to build conventional transmitter and receiver processing chains.

[0002] Although transmitting and receiving DNN models are being considered for deployment in advanced communications systems, conventional approaches often do not provide for sufficient control of DNN models to rapidly generate outputs resulting in whitened physical transmission signals for reducing interference when necessary.

[0003] For example, although a transmitting DNN model may be trained to generate an output communication signal for whitening the transmission signal, it may not be practical to maintain a particular white noise interference level of the transmission signal (e.g., an amount of white noise interference that is tolerated by the system or a particular power spectral density (PSD) level of a white noise spectrum that is tolerated) for the different combinations of input communication data that may be processed by the transmitting DNN model.

[0004] There is an opportunity to develop an efficient and dynamic mechanism capable of controlling a transmitting DNN model for adjusting a transmission signal, which when transmitted, is whitened or looks random to neighbor cells (e.g., whitened interference), and/or meets a certain white noise interference level while still maintaining the advantages of end-to-end communications using transmitting and receiving DNNs.

SUMMARY

[0005] In a first aspect, the present disclosure provides a method performed by a first device in communication with a second device, the method comprising: processing input communication data with a transmitting deep neural network (DNN) for generating an output communication signal for transmission to the second device; performing a scrambling DNN operation in response to a forecasted transmission of the generated output communication signal not satisfying a white noise interference level, said scrambling DNN operation further comprising: selecting neural network scrambling information (NNSI) for reconfiguring the transmitting DNN to process the input communication data to generate a scrambled output communication signal, which when transmitted satisfies the white noise interference level; transmitting, to the second device, a control message indicating the NNSI and scrambling timing information for directing when the second device is to reconfigure a receiving DNN; and transmitting, to the second device, the scrambled output communication signal that satisfies the white noise interference level based on the scrambling timing information.

[0006] In a second aspect, the present disclosure provides a method performed by a second device in communication with a first device, the method comprising: receiving, from the first device, a control message indicating NNSI and scrambling timing information; receiving, from the first device, a communication signal transmitted according to the scrambling timing information; reconfiguring a receiving DNN of the second device according to the NNSI and the scrambling timing information; processing the received communication signal with the receiving DNN for generating reconstructed communication data represented by the received communication signal; and sending the reconstructed communication data to a data sink of the second device or sending the reconstructed communication data to one or more upper protocol layers of a protocol stack of the second device.

[0007] Further aspects provide apparatus and systems for implementing the methods of the first and second aspects.

[0008] Aspects of the methods, apparatus and systems provide numerous advantages including, for example, efficient design and control of transmitting and receiving DNN structures that are capable of maintaining a transmission signal that satisfies a white noise interference level and for reducing interference to other DNN or non-DNN receivers in the cell or region around the first and second devices. Scrambling DNN operations used on the transmitting DNN of the first device maintain a white noise interference level of transmission signals from the first device without generating accidental transmission spikes due to the multiplicity of different combinations of input communication data processed by the transmitting DNN. The scrambling DNN operations mitigate, reduce and/or prevent transmission spikes in transmission signals of the first device from occurring when using the transmitting DNN whilst satisfying a white noise interference level. A further advantage includes efficiently controlling the reconfiguration of the receiving DNN of the second device due to corresponding scrambling DNN operations used on the transmitting DNN of the first device. The transmitting and receiving DNNs of the first and second device are reconfigured efficiently, rapidly, and dynamically according to the scrambling DNN operations in real-time to change the white noise interference level of transmission signals generated using the output communication signal of a transmitting DNN, while at the same time maintaining the transmission power or bit I symbol error rate of the signal of interest. A further advantage includes the efficient synchronisation between a first device using a transmitting DNN and a second device using a corresponding receiving DNN to enable dynamic whitening of the transmission signal from the first device and enable reception and decoding of the dynamically whitened transmission signal by the second device.

BRIEF DESCRIPTION OF THE DRAWINGS

[0009] The present disclosure may be better understood, and its numerous features and advantages made apparent to those skilled in the art by referencing the accompanying drawings. The use of the same reference symbols in different drawings indicates similar or identical items. Embodiments of the invention will be described, by way of example, with reference to the following drawings, in which.

[0010] FIG. 1 is a schematic diagram illustrating a comparison between example conventional transmitter and receiver structures and example end-to-end communication transmitter and receiver structures using deep neural networks in accordance with some embodiments.

[0011] FIG. 2a is a schematic diagram illustrating an example scrambling DNN communication system in accordance with some embodiments;

[0012] FIG. 2b is a schematic diagram illustrating an example power spectral density of a transmission signal satisfying a white noise interference level in accordance with some embodiments;

[0013] FIG. 2c is a schematic diagram illustrating another example power spectral density of a transmission signal with transmission spikes not satisfying a white noise interference level in accordance with some embodiments;

[0014] FIG. 2d is a schematic diagram illustrating a further example power spectral density of another transmission signal exceeding a white noise interference level in accordance with some embodiments;

[0015] FIG. 2e is a schematic diagram illustrating a further example power spectral density of yet another transmission signal satisfying a white noise interference level in accordance with some embodiments;

[0016] FIG. 3a is a schematic diagram illustrating an example input-based scrambling configuration for a transmitting and receiving DNN in accordance with some embodiments;

[0017] FIG. 3b is a schematic diagram illustrating an example output-based scrambling configuration for a transmitting and receiving DNN in accordance with some embodiments; [0018] FIG. 3c is a schematic diagram illustrating an example hidden-layer-based scrambling configuration for a transmitting and receiving DNN in accordance with some embodiments;

[0019] FIG. 4a is a flow diagram illustrating an example DNN scrambling process for generating output communication signal from a transmitting DNN that satisfies a white noise interference level when transmitted in accordance with some embodiments;

[0020] FIG. 4b is a flow diagram illustrating an example process for analysing whether the spectral density of a transmission signal representing the output communication signal from a transmitting DNN satisfies a white noise interference level in accordance with some embodiments;

[0021] FIG. 4c is a flow diagram illustrating an example process for receiving one or more control messages including neural network scrambling information at a second device in accordance with some embodiments;

[0022] FIG. 4d is a flow diagram illustrating an example process for receiving, at a second device, a transmission of output communication signal from a transmitting DNN of a first device and reconstructing the communication data in accordance with some embodiments;

[0023] FIG. 5 is a schematic diagram illustrating an example first device transmitter with a transmission buffer and an example second device receiver with a receiving buffer in accordance with some embodiments;

[0024] FIG. 6a is a schematic diagram illustrating an example random permutation and random inverse permutation (or depermutation) for use in scrambling one or more neural network layers of a transmitting DNN in accordance with some embodiments;

[0025] FIG. 6b is a schematic diagram illustrating an example random permutation sequence starting from an initial seed in accordance with some embodiments;

[0026] FIG. 6c is a schematic diagram illustrating example permutation matrices resulting from a selected i-th random permutation sequence for use in scrambling / descrambling one or more neural network layers of a transmitting/receiving DNN in accordance with some embodiments;

[0027] FIG. 6d is a flow diagram illustrating an example iterative process for selecting an i-th random permutation sequence for use in randomizing the order of neural network nodes of one or more neural network layers of a transmitting DNN, so that transmission of the output communication signal satisfies a white noise interference level in accordance with some embodiments;

[0028] FIG. 7 is a signal flow diagram illustrating enabling and disabling a scrambling DNN communications session between a first device and a second device in accordance with some embodiments;

[0029] FIG. 8 is a signal flow diagram illustrating example DNN scrambling communications between the first device and second device during the DNN communications session illustrated in FIG. 7 in accordance with some embodiments;

[0030] FIG. 9 is a signal flow diagram illustrating another example DNN scrambling communications between the first device and second device during the DNN communications session illustrated in FIG. 7 in accordance with some embodiments;

[0031] FIG. 10 is a signal flow diagram illustrating a scrambling DNN communications session between a first device, a second device and a third device in accordance with some embodiments;

[0032] FIG. 11 is a signal flow diagram illustrating example DNN scrambling communications between the first device, second device and a third device during the DNN communications session illustrated in FIGs. 7 or 10 in accordance with some embodiments;

[0033] FIG. 12 is a signal flow diagram illustrating enabling and disabling an uplink (UL) I downlink (DL) scrambling DNN communications session between a base station and a user equipment in accordance with some embodiments;

[0034] FIG. 13 is a signal flow diagram illustrating example DL DNN scrambling communications between the base station and user equipment during the UL / DL DNN communications session of FIG. 12 in accordance with some embodiments; [0035] FIG. 14 is a signal flow diagram illustrating example UL DNN scrambling communications between the user equipment and base station during the UL I DL DNN communications session of FIG. 12 in accordance with some embodiments;

[0036] FIG. 15 is a signal flow diagram illustrating another example UL DNN scrambling communications between the user equipment and base station during the UL / DL DNN communications session of FIG. 12 in accordance with some embodiments;

[0037] FIG. 16 is a signal flow diagram illustrating another example DNN scrambling communication session between a base station and a user equipment in accordance with some embodiments;

[0038] FIG. 17 is a schematic diagram of an example computer-readable medium in accordance with some embodiments.

DETAILED DESCRIPTION

[0039] FIG. 1 illustrates a comparison between example conventional transmitter and receiver structures for first and second conventional communication devices 102a and 102b and an example end-to-end communication transmitter and receiver structures for first and second deep neural network (DNN) devices 104a and 104b (also referred to herein as first and second devices 104a and 104b) using transmitting and receiving DNN structures 106 and 108 (TX DNN and RX DNN), respectively. Conventional 4G and 5G communication systems have complicated transmitter and receiver processing chains with multiple processing components for encoding and modulating input communication data 101a for wireless transmission from the first conventional device 102a and reception as reconstructed communication data 101b by the second conventional communication device 102b.

[0040] In this example, a transmitter processing chain of the first conventional communication device 102a processes input communication data 101 a (e.g., a block of bits, a bit stream or other digital data) from a data source (not shown) for transmission from the first conventional communication device 102a to the second conventional communication device 102b. The transmitter processing chain includes an arrangement of processing blocks such as, for example, an encoding block, an interleaving block, a scrambling block, a pre-coding block, and a modulation block, which process the input communication data 101a into an output communication signal 118 for transmission. The first communication device 102a includes a radio frequency (RF) front-end that includes RF Analog transmission/transmitter/ transmitting (TX) components (RF Analog TX) 103a for processing the output data signal (e.g., Digital-to-Analog conversion and/or radio frequency upconversion etc.) for radio frequency upconversion and transmitting via antennas as a transmission signal 105a over communication channel 105. The second conventional communication device 102b receives the transmission signal 105a. The second communication device 102b includes a RF front-end that includes RF Analog reception/receiver/receiving (RX) components (RF Analog RX) 103b of the second conventional communication device 102b for receiving the transmission signal 105a (e.g., analog-to-digital conversion and/or frequency down conversion to baseband) and the receiver processing chain generates reconstructed communication data 101b from the resulting received transmission signal 105a. The receiver processing chain includes an arrangement of processing blocks such as, for example, demodulation, descrambling, de-interleaving, and decoding components for processing the received transmission signal 105a and to recover the input communication data 101a as reconstructed communication data 101 b. As communications systems evolve (e.g., 5G to 6G communications standards) to provide higher capacity and lower latency along with convergence of various disparate technologies, the already complex transmitter and receiver processing chains will experience various updates and modifications. This will result in regimented, complex transmitter and receiver processing chains.

[0041] The complexity of current transmitter and receiver processing chains can be reduced by training machine learning (ML) algorithms, such as deep neural networks (DNNs), to integrate transmitter and receiver chain processing blocks in the form of transmitting DNN and receiving DNN structures 106 and 108 (also referred to herein as a transmitter DNN structure and receiver DNN structure, respectively) capable of providing end-to-end communications. Such transmitting and receiving DNN structures 106 and 108 potentially augment and/or replace conventional transmitter and receiver processing chains used in first and second conventional communication devices 102a and 102b. For example, a transmitting DNN structure 106 includes a transmitting DNN model 107 (TX DNN) trained to replace the transmitter processing chain including the encoding, interleaving, scrambling, and modulation blocks of the first conventional device 102a. The first device 104a configures the transmitting DNN structure 106 to process input communication data 101a’ (e.g., a block of input communication data I bits) and generate an output communication signal 118 for transmitting, via RF Analog TX 103a’, as a transmission signal 105a’ over communication channel 105’. In such a case, the transmitting DNN model 107 of the transmitting DNN structure 106, when trained, generates the output communication signal 118. The RF Analog TX 103a’ processes the output communication signal 118 for transmitting as the transmission signal 105a’ over communication channel 105’. The resulting transmission signal 105a’ generated from the output communication signal 118 of the transmitting DNN model 107 is a transmission waveform suited to, depending on the training /conditions, efficiently address numerous channel environments, impairments, and interference found in current communication systems (e.g., multi-path interference, multiple access interference, narrowband interference).

[0042] The second device 104b receives the transmission signal 105a’ via RF Analog RX 103b’, which processes the received transmission signal 105a’ (e.g., performs at least down conversion) into a baseband received communication signal 119 for input to a receiving DNN structure 108. The receiving DNN structure 108 includes a receiving DNN model 109 (RX DNN) trained to generate reconstructed communication data 101b’ when given, as input, a suitably formatted baseband (or down-converted) input data signal. The second device 104b configures the receiving DNN model 109 to perform the reciprocal operations of the transmitting DNN model 107 for generating reconstructed communication data 101b’ that is representative of the input communication data 101a’ input to the transmitting DNN model 107.

[0043] In an example, the second device 104b has a protocol stack with a plurality of protocol layers. After generating reconstructed communication data 101 b’ for a particular time slot or for one or more time slots, the second device 104b sends the reconstructed communication data 101 b’ for each one or more time slots to one or more upper layer protocols of the protocol stack of the second device 104b. In the protocol stack of the second device 104b, the lower layers are responsible for providing services to the upper layers, and the upper layers use those services to provide their own functions. For example, the reconstructed communication data 101b’ is generated at the physical layer of a protocol stack and passed up and processed by each of the upper layers until the application layer of the protocol stack, where the corresponding reconstructed communication data 101 b’ is used for, without limitation, for example display to a user, further processing, and/or sending to one or more applications of the second device 104b for further processing and/or consumption of the reconstructed communication data 101 b’.

[0044] In this example, the receiving DNN model 109, when trained, replaces the receiver processing chain including, for example, the demodulation, descrambling, deinterleaving, decoding blocks for generating the reconstructed communication data 101b’ from the received transmission signal 105a’. The reconstructed communication data 101b’ is representative of the input communication data 101a’. The transmitting and receiving DNN models 107 and 109 are well suited for supporting end-to-end communication systems where it may be impractical to build conventional transmitter and receiver processing chains. Transmitting and receiving DNN structures 106 and 108 will become critical components of 5G advanced or even 6G and beyond communications systems as transmitter and receiver chain complexities and requirements increase.

[0045] Although the transmitting DNN model 107 is described as performing the functions of encoding, interleaving, scrambling, and modulation blocks of a transmitter chain, this is by way of example only and it is not so limited. It is to be appreciated by the skilled person that the transmitting DNN model 107 is trained to perform any one or more functions of a transmitter processing chain that includes at least one or more of encoding, interleaving, scrambling, precoding, modulation and the like, combinations thereof, modifications thereto and/or as the application demands. Although the receiving DNN model 109 is described as being trained to perform the functions of the demodulation, descrambling, deinterleaving, decoding of a receiver processing chain, this is by way of example only and it is not so limited. It is to be appreciated by the skilled person that the receiving DNN model 109 is trained to perform one or more functions of a receiver processing chain that includes at least one or more of demodulation, descrambling, deinterleaving, decoding and/or any other receiver processing chain function and the like, combinations thereof, modifications thereto and/or as the application demands. For example, the transmitting DNN model 107 is trained to perform many or most, if not all, the functions of a transmit processing chain apart from the modulation block and RF Analog TX 103a’, where the corresponding receiving DNN model is trained to perform most, if not all, the functions of a receiver processing chain apart from the RF Analog RX 103b’ and demodulation block.

[0046] Each transmitting and receiving DNN model 107 and 109 of the corresponding transmitting and receiving DNN structure 106 and 108, respectively, as described herein have been trained to replace the functions of the conventional transmitting/receiving chains, respectively, and/or for overcoming various channel conditions, and/or to meet the performance requirements of 5G, 6G and future communications standards and the like. The transmitting and receiving DNN models 107 and 109 are deployable for configuring the transmitting and receiving DNN structures 106 and 108, respectively, for use in performing DNN communications between the first device 104a and the second device 104b, respectively. For example, the first DNN model 107 includes at least an input neural network layer, one or more hidden neural network layers, and an output neural network layer. The first DNN model 107 processes input communication data and generates an output communication signal for processing and transmission by RF Analog TX 103a’ (e.g., Digital-to-Analog conversion and radio frequency upconversion etc.) as a transmission signal 105a’. For example, the second DNN model 109 includes at least an input neural network layer, one or more hidden neural network layers, and an output neural network layer. The RF Analog RX 103b’ processes the transmission signal 105a’ (e.g., analog-to-digital conversion and/or frequency down conversion to baseband) to generate a received communication signal 119 for input to the second DNN model 109, which processes the received communication signal to generate reconstructed communication data representing the input communication data incorporated in the transmission signal 105a’. In another example, a pair of first and/or second DNN models 107 and/or 109 can be selected by the first device 104a depending on communication performance requirements and the like for a DNN communication session with the second device 104b, the selection of which the first device 104a can communicate to the second device 104b during establishment of the DNN communication session.

[0047] The ML model algorithms/architectures used to train the transmitting and receiving DNN model 107 and 109 as described herein are based on or include, by way of example only but is not limited to, one or more of: a neural network, a fully connected neural network, a convolutional neural network, a long short-term memory (LSTM) neural network, and a transformer neural network, and/or any other suitable DNN architecture, combinations thereof, modifications thereto, as herein described, and/or as the application demands Supervised and/or unsupervised training may be performed as the application demands. For example, supervised training of the first and second DNN models 107 and 109 for use by first and second devices 104a and 104b may use, for example, gradient back-propagation based techniques for updating the weights I parameters of, for example, nodes of the neural network layers and any other DNN architecture components of the corresponding first and/or second DNN models 107 and/or 109 using, for example, an appropriate or suitable loss function. It is assumed that the various DNN models and/or architectures and the DNN model/structure arrangements as described with reference to FIGs. 3a to 3c for the first and second DNN structures of the first and second devices have already been trained and determined. A plurality of different first and/or second DNN models 107 and 109 for use in different communication scenarios can be stored and/or mapped with suitable identifiers I indexes in storage accessible to first and/or second devices 104a and 104b for retrieval to configure the first and second DNN structures 106 and 108 of the first and second devices 104a and 104b when first and second devices 104a and 104b establish DNN communications therebetween.

[0048] Although there are significant advantages of using transmitting and receiving DNN structures 106 and 108 for replacing one or more functionalities of conventional transmitting and receiving chains, there are also some issues to consider when implementing transmitting and receiving DNN structures 106 and 108 to meet performance requirements of 5G, 6G and future communications standards. For example, rapidly generating a whitened physical transmission signal that reduces interference or that meets a certain white noise interference level is beneficial for increasing capacity and reducing latency in communication systems. Currently, the scrambling/descrambling blocks of conventional 4G/5G transmitter and receiver chains are individually controlled to rapidly whiten a physical transmission signal. Such functionality is difficult to incorporate into current transmitting and receiving DNN structures though. Although the transmitting and receiving DNN models 107 and 109 of the transmitting and receiving DNN structures 106 and 108 can be trained together (e.g., jointly) to generate an output communication signal 118, which when upconverted and transmitted as a transmission signal 105a’, meets a certain white interference noise level (e.g., an amount of white noise interference that is tolerated by the system or a particular PSD level of a white noise spectrum that is tolerated), it may not be possible to maintain this particular white noise interference level for subsequent input communication data blocks due to changes in the input bitstream.

[0049] In addition, different receivers in a cell or region can tolerate different levels of interference, where a transmitter chain of the first device 104a dynamically adjusts the white noise interference level of the transmission signal. For transmitting DNN structures 106 that have replaced a conventional transmitter processing chain, then those remaining components such as, for example the RF Analog TX 103a’, dynamically adjust the final transmission signal to meet white noise interference levels. However, using RF Analog TX 103a’ results in a coarse adjustment of transmission power and risks increased symbol or bit error rates over the communication link with a subsequent reduction in throughput due to, for example, increased retransmissions between first and second devices 104a and 104b.

Another possible approach is to train multiple transmitting DNN models (and corresponding receiving DNN models) each configured to generate an output communication signal 118, which when transmitted by the RF Analog TX 103a’, that satisfies a different white noise interference level for the same input communication data 101a’. The first device 104a selects the transmitting DNN model (and corresponding receiving DNN model) that generates the output communication signal 118 with lowest white noise interference level forecasted for transmission. This may be infeasible due to the sheer number of transmitting DNN models and corresponding receiving DNN models to meet all plausible white interference noise levels. This is also an inefficient and impractical use of computational resources at both first and second devices 104a and 104b.

[0050] Training multiple different transmitting DNN models (and corresponding receiving DNN models) for use in whitening the transmission signal 105a’ to satisfy different white noise interference levels, where depending on the white noise interference level a transmitting and receiving DNN pair is selected for use by the transmitting and receiving DNN structures 106 and 108 of the first and second devices 104a and 104b, respectively, is a resource intensive process requiring a large amount of computing, storage and transmission resources to reliably cater for all types of interference and white noise interference levels. This means the first and second device have multiple transmitting and receiving DNN structures stored thereon to call upon.

[0051] The above issues are addressed by including a scrambling operation into the transmitting DNN structure 106 of the first device 104a that is controllable for generating an output communication signal 118 from the input communication data 101a’, which when processed and transmitted by the RF Analog TX 103a’, generates a transmission signal 105a’ satisfying the white noise interference level. The receiving DNN structure 108 of the second device 104a performs descrambling using reciprocal operations of the transmitting DNN structure 106 to generate reconstructed communication data 101b’. The scrambling DNN operations are controllable using neural network scrambling information (NNSI). The first device 104a selects the NNSI from a set of NNSI, each NNSI associated with a different white noise interference level. For example, each NNSI describes the type of scrambling and/or where the scrambling occurs within one or more neural network layers of the transmitting DNN model 107 of the transmitting DNN structure 106. The first device 104a selects an NNSI from the set of NNSI to scramble the output communication signal 118 when the first device 104a forecasts that the resulting transmission signal 105a’ satisfies a particular white noise interference level. The first device 104a transmits the selected NNSI to the second device 104b, prior to transmitting the corresponding output communication signal 118 as transmission signal 105a’. This enables the second device 104b to reconfigure the receiving DNN structure 108 for generating reconstructed communication data 104b’ corresponding to the input communication data 101a’ represented by the transmission signal 105a’.

[0052] For example, when the first device 104a communicates with the second device 104b, the transmitting DNN structure 106 of the first device 104a processes the input communication data 101a’ using the transmitting DNN model 107 of the transmitting DNN structure 106 for generating an output communication signal 118 for transmission to the second device 104b. When the first device 104a analyzes the generated output communication signal 118 and estimates or forecasts the resulting transmission would not satisfy a white noise interference level, then, in response, the first device 104a and second device 104b perform a scrambling DNN operation. The scrambling DNN operation at the first device 104a includes the first device 104a selecting an NNSI for reconfiguring the transmitting DNN model 107 of the transmitting DNN structure 106 to process the input communication data 101a’ and generate a scrambled output communication signal 118, which satisfies the white noise interference level when transmitted as transmission signal 105a’. First device 104a transmits the selected NNSI to the second device 104b in a control message. The control message includes an indication of the NNSI and scrambling timing information. The scrambling timing information is for directing when the second device 104b reconfigures the receiving DNN model 109 of the receiving DNN structure 108 for generating the reconstructed communication data 101 b’ when it receives the transmission signal 105a’ corresponding to the scrambled output communication signal 118. At the appropriate time, based on the scrambling timing information, the first device 104a processes the scrambled output communication signal 118 via the RF Analog TX components as transmission signal 105a’ to the second device 104b, where transmission of the scrambled output communication signal 118 satisfies the white noise interference level.

[0053] The scrambling DNN operation at the second device 104b includes the second device 104b receiving the control message indicating NNSI and the corresponding scrambling timing information. The RF Analog RX 103b’ of the second device 104b receives according to the scrambling timing information, from the first device, the transmission signal 105a’ from the communication channel 105 and outputs a received communication signal 119 for processing by the receiving DNN model 109 of the second device 104b. The second device 104b, prior to processing the received communication signal 119, reconfigures the receiving DNN model 109 using the received NNSI and associated scrambling timing information. After reconfiguration, the receiving DNN model 109 processes the received communication signal 119 and generates reconstructed communication data 101b’ represented by the received communication signal 119. The second device 104b sends the reconstructed communication data 101 b’ to a data sink of the second device 104b or sends the reconstructed communication data 101b’ to one or more upper protocol layers of a protocol stack of the second device 104b (e.g., to an application protocol layer of a protocol stack for use by one or more applications executing on the second device 104b). The scrambling DNN operation continues while the transmission signal 105a’ transmitted by the first device 104a satisfies the white noise interference level and/or there is additional input communication data for transmission.

[0054] During the communication session, when the first device 104a performs multiple scrambling DNN operations, the first device 104a selects different NNSI because different input communication data 101a’ causes the transmitting DNN model 107 to generate an output communication signal 118, which when transmitted as a transmission signal 105a’ by RF Analog TX 103a’, will not satisfy the current white noise interference level. The first device 104a sends the different selected NNSI in control messages to the second device 104b along with associated scrambling timing information for use in the corresponding scrambling DNN operation. The first device 104a selects different NNSI when the white noise interference level changes and, as a result, a victim device may experience intolerable interference from the first device 104a or the first device 104a receives requests for adjustment of the white noise interference level to a tolerable interference level.

[0055] The scrambling DNN operations performed by the first and second devices 104a and 104b provide the advantage that the first device 104a does not transmit the output communication signal 118 of the transmitting DNN until the resulting transmission signal satisfies the white noise interference level set by the first device 104a. This means that transmissions of the first device 104a will satisfy the white noise interference level without transmission spikes causing interference to neighboring devices. The scrambling DNN operations also synchronizes the transmitting DNN structure 106 and receiving DNN structure 108 to operate together to recover the input communication data 101a’ at the second device 104b.

[0056] The first and second devices 104a and 104b can be any type of communication device for use in communication system 100 such as, but not limited to, for example any combination of radio access network (RAN) elements including a base station (BS), network devices, user equipment (UE), or other RAN elements within communication system 100. For example, the first device 104a and second device 104b may be two BSs, or two UEs, or a BS and a UE, or a UE and a BS, or any other combination of communication devices as the application demands. FIG. 2a illustrates a scrambling DNN communication system 200 in which the first device 210 and second device 220 are a BS and a UE, respectively.

[0057]

[0058] FIG. 2a illustrates an example scrambling DNN communication system 200 in which a first device 210 is in communication with a second device 220. In this case, the first device 210 is a BS and the second device 220 is a UE. The BS 210 connects via one or more interfaces to a core network (not shown) of the scrambling DNN communication system 200. For example, the communication system may be a 5G / 6G or new radio (NR) communication system. The UE 220 and BS 210 communicate via downlink transmission signal 205a (e.g., downlink transmissions) and uplink transmission signal 205b (e.g., uplink transmissions) over a wireless communication channel 205. The wireless communication channel 205 may include a downlink communication channel (e.g., Physical Downlink Shared Channel (PDSCH)) over which a BS 210 transmits a downlink transmission signal 205a to UE 220 and an uplink communication channel (e.g., Physical Uplink Shared Channel (PUSCH)) over which the UE 220 transmits an uplink transmission signal 205b to the BS 210. The downlink communication channel may also include a downlink control channel (e.g., Physical Downlink Control Channel (PDCCH)) and the uplink communication channel may also include an uplink control channel (e.g., Physical Uplink Control Channel (PUCCH)).

[0059] The BS 210 may be implemented as a computing system/apparatus for performing any of the corresponding methods, scrambling DNN operations, scrambling/randomizing/descrambling operations or processes described herein and/or for implementing any of the corresponding systems, units and/or apparatus as described herein. The BS 210 includes a RF front-end 203a/b including RF Analog TX and RX components/antennas etc., one or more transceivers 211 , one or more processors 212, and a memory unit 213 connected together. It will be appreciated by the skilled person that other types of computing devices/systems/platforms may alternatively be used to implement the BS 210 and the methods described herein, such as a distributed computing system as the application demands. The BS 210 includes one or more processors 212. The one or more processors 212 control operation of other components of the BS 210 such as RF front-end 203a/b, one or more transceivers 211 , the memory unit 213 and the like. The one or more processors 212 may be a single core device or a multiple core device. The one or more processors 212 may comprise a Central Processing Unit (CPU), one or more CPUs, a graphical processing unit (GPU), and/or one or more GPUs and the like. Alternatively, the one or more processors 212 may comprise specialized processing hardware, for instance a reduced instruction set computer (RISC) processor or programmable hardware with embedded firmware. Multiple processors may be included in BS 210. In some embodiments, the one or more processors 212 may be part of a distributed computing system such as a cloud computing system and/or cloud computing platform.

[0060] The one or more processors 212 of the BS 210 may be connected to a network interface such as, for example, transceivers 211 including a transmitter (TX) and a receiver (RX) for communicating via RF Front end 203a/b over wireless communication channel 205 of a network with other apparatus and systems such as UE 220, other communication devices, network equipment, RAN entities or devices, operators and/or any other apparatus, service, system and/or device as the application demands. The one or more processors 212 may, optionally, be connected with a user interface (Ul) for user or operator input for instructing or using the BS 210 and/or underlying computing system and/or for outputting data therefrom. The one or more processors 212 may, optionally, be connected with a display for displaying output to a user or operator.

[0061] The BS 210 includes memory system or memory unit 213 including a working or volatile memory. The one or more processors 212 may access the volatile memory in order to process data and may control the storage of data in memory.

The volatile memory may comprise random access memory (RAM) of any type, for example, Static RAM (SRAM), Dynamic RAM (DRAM), or it may comprise Flash memory, such as an Secure Digital (SD)-Card. In some embodiments, the memory unit 213 and/or one or more volatile memories may comprise a multiple of a plurality of memory forming part of the distributed computing system such as the cloud computing system and/or cloud computing platform and the like. The BS 210 also includes a non-volatile memory. The non-volatile memory may store a set of operation or operating system instructions for controlling the operation of the processors 212 in the form of computer readable instructions and/or software instructions in the form of computer readable instructions, which when executed on the one or more processors cause the processors to implement the methods, processes, operations and/or functionality of the scrambling DNN operations, scrambling/randomizing operations, processes and/or methods as described herein. The non-volatile memory may be a memory of any kind such as a read only memory (ROM), a Flash memory, SD drive, a magnetic drive memory or magnetic disc drive memory and the like as the application demands. In some embodiments, the nonvolatile memory may comprise a multiple of a plurality of non-volatile memory forming part of the distributed computing system such as the cloud computing system and/or cloud computing platform and the like.

[0062] The non-volatile memory of the memory unit 213 of the BS 210 includes computer program code and/or instructions for implementing a BS DNN Controller (DNNC) 214, and/or a BS downlink transmitting DNN structure (BS DL TX DNN) 206, or BS uplink receiving DNN structure (BS UL RX DNN) 208. The BS DNNC 214, when executed on the one or more processors 212, controls scrambling DNN operations between the BS 210 and the UE 220 using a BS DL TX DNN 206 and/or BS UL RX DNN 208, a BS neural network scrambling information (NNSI) store / table I buffer 215a stored in memory unit 213, and BS transmitting/receiving (TX/RX) DNN store or table 215b (also referred to as BS TX DNN / RX DNN store or table 215b) stored in memory unit 213. Although the BS DNNC 214 is illustrated as part of the memory unit 213, this is by way of example only and the BS DNNC 214 is not so limited. It is to be appreciated by the skilled person that the BS DNNC 214 may be implemented in hardware and/or software of the BS 210 as the application demands.

[0063] The at least one processor 212, with the at least one memory unit 213 and computer program code or instructions stored thereon are arranged to cause the computing system of the BS 210 to at least perform at least the corresponding operations, methods, and/or processes, for example as disclosed in relation to the schematic diagrams, flow diagrams or operations as described with any of FIGs. 1 to 17 and related features thereof.

[0064] Similarly, the UE 220 may be implemented as a computing system/ apparatus for performing any of the corresponding methods, scrambling DNN operations, scrambling/randomizing/descrambling operations or processes described herein and/or for implementing any of the corresponding systems, units and/or apparatus as described herein. The UE 220 includes an RF front-end 203a/b, one or more transceivers 221 , one or more processors 222, and a memory unit 223 connected together. It will be appreciated by the skilled person that other types of computing devices/systems/platforms may alternatively be used to implement the UE 220 and the methods described herein. The UE 220 includes one or more processors 222 (e.g., CPUs). The one or more processors 222 control operation of other components of the UE 220 such as RF front-end 203a/b, one or more transceivers 221 , the memory unit 223 and the like. The one or more processors 222 may be a single core device or a multiple core device. The one or more processors 222 may comprise a CPU and/or a GPU. Alternatively, the one or more processors 222 may comprise specialized processing hardware, for instance a RISC processor or programmable hardware with embedded firmware. Multiple processors may be included in UE 220.

[0065] The one or more processors 222 of the UE 220 may be connected to a network interface such as, for example, transceivers 221 including a transmitter (TX) and a receiver (RX) for communicating via RF front-end 203a/b over wireless communication channel 205 of a network with other apparatus and systems such as BS 210, other communication devices, network equipment, RAN entities or devices, users or operators and/or any other apparatus, service, system and/or device as the application demands. The one or more processors 222 may, optionally, be connected with a Ul for user input for instructing or using the UE 220 and/or underlying computing system and/or for outputting data therefrom. The one or more processors 222 may, optionally, be connected with a display for displaying output to a user.

[0066] The UE 220 includes memory system or memory unit 223 including a working or volatile memory. The one or more processors 222 may access the volatile memory in order to process data and may control the storage of data in memory.

The volatile memory may comprise RAM of any type, for example, SRAM, DRAM, or it may comprise Flash memory, such as an SD-Card. The UE 220 also includes a non-volatile memory. The non-volatile memory may store a set of operations or operating system instructions for controlling the operation of the processors 222 in the form of computer readable instructions and/or software instructions in the form of computer readable instructions, which when executed on the one or more processors 222 cause the processors 222 to implement the corresponding methods, processes, operations and/or functionality of the scrambling DNN operations, scrambling/randomizing/ descrambling operations and/or methods at the UE 220 as described herein. The non-volatile memory may be a memory of any kind such as a ROM, a Flash memory, SD drive, a magnetic drive memory or magnetic disc drive memory and the like as the application demands.

[0067] The non-volatile memory of memory unit 223 of the UE 220 includes computer program code and/or instructions for implementing a UE DNN Controller (UE DNNC) 224 and/or a UE uplink transmitting DNN structure (UE UL TX DNN) 226, and/or UE downlink receiving DNN structure (UE DL RX DNN) 228. The UE DNNC 224, when executed on the one or more processors 212, controls scrambling DNN operations between the BS 210 and the UE 220 using a UE UL TX DNN 226 and/or UE DL RX DNN 228, a UE NNSI store I table I buffer 225a stored in memory unit 223, and UE TX/RX DNN store I table 225b stored in memory unit 223. Although the UE DNNC 224 is illustrated as part of the memory unit 223, this is by way of example only and the UE DNNC 224 is not so limited. It is to be appreciated by the skilled person that the UE DNNC 224 may be implemented in any combination of hardware and/or software of the UE 220 and/or as the application demands.

[0068] In operation, the BS 210 and UE 220 establish a DL/UL DNN communication session between each other. The DL/UL DNN communication session includes DL DNN communications from BS 210 to UE 220, and UL DNN communications from UE 220 to BS 210. During establishment of the DL/UL DNN communication session, the BS 210 and UE 220 communicate with each other for defining, agreeing, and/or configuring the type of BS DL TX DNN 206 and UE DL RX DNN 228 pair for use in DL DNN communications, and the type of UE UL TX DNN 226 and BS uplink receiving DNN structure (BS UL RX DNN) 208 pair for use in UL DNN communications. For example, the BS 210 selects an appropriate BS DL TX DNN 206 and UE DL RX DNN 228 pair for DL DNN communications from BS TX DNN/RX DNN store or table 215b (e.g., a DNN configuration table). Similarly, the BS 210 selects an appropriate UE UL TX DNN 226 and BS UL RX DNN 208 pair for UL DNN communications from BS TX DNN/RX DNN store or table 215b. Alternatively, as an option, the UE 220 might select an appropriate UE UL TX DNN 226 and BS UL RX DNN 208 pair for UL DNN communications from UE TX DNN/RX DNN store 225b.

[0069] The BS TX/RX DNN store or table 215b includes a set of DL transmitting/receiving DNN structures and/or pairs thereof and a set of UL transmitting/receiving DNN structures and/or pairs thereof each of, which are mapped to DL I UL DNN identifiers, respectively, and stored in the BS TX/RX DNN store or table 215b (e.g., a look-up table) at the BS 210. Each transmitting/receiver DNN structure is trained to transmit/receive a transmission waveform suited to, depending on the training/conditions, efficiently overcome, for example, a particular channel environment, one or more particular channel impairments, and/or one or more different types of interference found in current communication systems (e.g., multipath interference, multiple access interference, narrowband interference) further enhancing performance. Training the DL/UL TX DNN and corresponding DL/UL RX DNN to overcome one or more types of channel impairments and/or channel environments use supervised DNN training over multiple scenarios. For example, the supervised DNN training jointly trains a pair of DL TX DNN I DL RX DNN and a pair of UL TX DNN I UL RX DNN. For example, the BS 210 and UE 220 use a DL transmitting DNN and receiving DNN pair trained specifically for downlink communication channels (e.g., PDSCH), and the UE 220 and BS 210 use an uplink transmitting DNN and receiving DNN pair trained specifically for uplink communication channels (e.g., PUSCH).

[0070] The UE 220 also has a corresponding set of UL I DL transmitting/receiving DNN structures also mapped to the same UL I DL DNN identifiers and stored in the UE TX DNN/RX DNN store 225b (e.g., a look-up table) at the UE 220. The BS 210 selects DL transmitting and receiving DNN structures and/or UL transmitting and receiving DNN structures based on the DL and/or UL communication channel I environment, the communication performance requirements for the DL and/or UL DNN connections, and the type of DL and/or UL data communications for the communication session (e.g., voice communication, data communications, multimedia streaming, and the like).

[0071] The BS 210 sends one or more control message indicating the selected

DL/UL receiving DNN structures (e.g., DL/UL DNN identifiers) for DL and UL communication channels (e.g., PDSCH and PUSCH), the DL and UL control channels (e.g., PDCCH/PUCCH) for use in DL and UL DNN communications between UE 220 and BS 210. For DL DNN communications, the BS 210 and UE 220 configure their respective BS DL TX DNN 206 and UE DL RX DNN 228 based on the selected DL DNN identifier. For UL DNN communications, the BS 210 and UE 220 configure their respective BS UL RX DNN 208 and UE UL TX DNN 226 based on the selected UL DNN identifier. After the BS 210 and UE 220 establish a DL/UL DNN communication session, the BS 210 and UE 220 perform DL DNN communications for one or more time slots using the BS DL TX DNN 206 and UE DL RX DNN 228. The UE 220 and BS 210 also perform UL DNN communications for one or more time slots using the UE UL TX DNN 226 and BS UL RX DNN 208, respectively.

[0072] During DL DNN communication, the BS 210 performs DL DNN scrambling when the BS 210 detects that the transmissions from the BS 210 do not satisfy a particular white noise interference level. During UL DNN communication, the UE 220 performs UL DNN scrambling when the UE 220 detects that the transmissions from the UE 220 does not satisfy a particular white noise interference level. For DL DNN scrambling communications from the BS 210 to the UE 220, a TX DNN controller of the BS DNNC 214 controls the DL DNN scrambling operations at the BS 210 and an RX DNN controller of the UE DNNC 224 controls the corresponding DL DNN scrambling operations at the UE 220. For UL DNN scrambling communications from the UE 220 to BS 210, a TX DNN controller of the UE DNNC 224 controls the UL DNN scrambling operations at the UE 220 and an RX DNN controller of the BS DNNC 214 controls the corresponding UL DNN scrambling operations at the BS 210.

[0073] For example, in DL DNN scrambling the BS 210 enables performance of the scrambling DL DNN operations for one or more time slots when an output communication signal from the BS DL TX DNN 206 does not satisfy the white noise interference level when forecast for transmission over PDSCH. The DL TX DNN controller of the BS 210 detects that the BS DL TX DNN 206 generates an output communication signal from input communication data for a specific time slot that would result in a downlink transmission signal 205a over the PDSCH that does not satisfy the white noise interference level. When this occurs, a DL scrambling DNN operation is enabled and the DL TX DNN controller of the BS 210 selects neural network scrambling information (NNSI), which may be retrieved from BS NNSI store 215a or iteratively determined by randomizing one or more neural network layers of the BS DL TX DNN 206, for use in reconfiguring the BS DL TX DNN 206 for generating an output communication signal from the same input communication data for the specific time slot that would result in a downlink transmission signal 205a over the PDSCH satisfying a white noise interference level set for the PDSCH. For example, selecting NNSI (e.g., randomization/scrambling parameters and/or specific/selected neural network layers of the transmitting DNN being randomized) for randomizing or scrambling the order of a set of network nodes of one or more specific neural network layers of the BS DL TX DNN 206 of the BS 210 such that the resulting output communication signal of the reconfigured BS DL TX DNN 206 produces a downlink transmission signal 205a that satisfies the white noise interference level for the PDSCH. The specific neural network layers of the BS DL TX DNN 206 that are randomized are specified in the NNSI.

[0074] Similarly, UL scrambling DNN operations are enabled when the UE 220 (or the BS 210) detects that the uplink transmission signal 205b over the PUSCH would not satisfy a white noise interference level set for the PUSCH. For example, the BS 210 selects NNSI for use in reconfiguring the UE UL TX DNN 226 of the UE 220 to generate an output communication signal therefrom that would result in an uplink transmission signal 205b over the PUSCH satisfying the white noise interference level. In an example, when the BS 210 selects the NNSI for the UE 220, the BS 210 simulates the UL with the random UE UL input communication data and UE UL TX DNN 226 to generate an output communication signal, and selects the NNSI that results in the output communication signal satisfying the white noise interference level for the UL. The BS 210 sends the selected NNSI to the UE 220 in a control message. The NNSI does not necessarily depend on the specific UE UL input communication data, rather it primarily depends on the UE UL TX DNN 226 used in the UE UL. For example, selecting NNSI (e.g., randomization/scrambling parameters and/or specific layers of the transmitting DNN being randomized) for randomizing or scrambling the order of a set of network nodes of one or more specific neural network layers of the UE UL TX DNN 226 of the UE 220 such that the resulting output communication signal of the reconfigured UE UL TX DNN 226 produces an uplink transmission signal 205b that satisfies the white noise interference level for the PUSCH. [0075] For DL scrambling communications, after selecting the NNSI for the BS DL TX DNN 206 that results in downlink transmission signal 205a satisfying the white noise interference condition, the BS 210 sends a control message to the UE 220 that includes an indication of the NNSI and scrambling timing information, which is used by the UE DNNC 224 to reconfigure the UE DL RX DNN 228 according to the scrambling timing information for processing the downlink transmission signal 205a resulting from the output communication signal generated from the reconfigured BS DL TX DNN 206. The received scrambling timing information directs when the UE DNNC 224 is to reconfigure the UE DL RX DNN 228 using the selected NNSI. For example, the UE DL RX DNN 228 processes the received downlink transmission signal 205a to generate reconstructed communication data for a specific time slot corresponding to input communication data transmitted for that specific time slot. As an example, the UE 220 sends the reconstructed communication data to a data sink of the UE 220 or sends the reconstructed communication data to one or more upper protocol layers of a protocol stack of the UE 220 (e.g., to an application protocol layer of a protocol stack for use by one or more applications executing on the UE 220).

[0076] Similarly, for UL scrambling communications, after selecting the NNSI for the UE UL TX DNN 226 that results in uplink transmission signals 205b satisfying the white noise interference condition, the BS 210 sends a control message to the UE 220 that includes an indication of the NNSI and scrambling timing information, which is used by the UE DNNC 224 to reconfigure the UE UL TX DNN 226 according to the scrambling timing information for processing the uplink transmission signals 205b for transmission over RUSCH to the BS 210, where the BS 210 receives the uplink transmission signals and processes using a reconfigured BS UL RX DNN 208 based on the selected NNSI and corresponding scrambling timing information. The scrambling timing information directs when the BS DNNC 214 is to reconfigure the BS DL TX DNN 208 using the selected NNSI. For example, the BS UL RX DNN 208 processes the received uplink transmission signals 205b to generate reconstructed communication data for a specific time slot corresponding to input communication data transmitted for that specific time slot.

[0077] Prior to transmitting the output communication signal generated by the BS DL TX DNN 206, the BS TX DNNC of the BS DNN Controller 214 performs an analysis of the output communication signal to identify whether the output communication signal will whiten the downlink transmission signal 205a and which satisfies a white noise interference level for PDSCH. When the analysis indicates RF processing of the output communication signal will result in generation of a transmission signal (e.g., a forecast transmit signal or forecast transmission signal) that does not satisfy the white interference noise level, then the BS 210 and UE 220 perform the above-mentioned selection of NNSI and scrambling DNN operations.

[0078] The DL/UL scrambling process provides the advantage that the BS 210 / UE 220 does not transmit the output communication signal of the DL/UL transmitting DNN until the resulting forecast downlink/uplink transmission signal 205a/205b satisfies the white noise interference level set by the BS 210 / UE 220. The DL/UL transmissions from the BS 210 and UE 220, respectively, will satisfy a corresponding white noise interference level without transmission spikes causing interference to neighboring devices and/or cells. The DL/UL scrambling process also synchronizes the DL/UL transmitting DNNs 206/226 and corresponding DL/UL receiving DNNs 228/208, respectively, to operate together and to recover the corresponding input communication data at the UE 220 or BS 210, respectively.

[0079] The DL/UL scrambling communications performed by the BS 210 and UE 220 provide numerous advantages including, for example, efficient design and control of BS DL TX DNN 206 and UE DL RX DNN 228, and or UE UL TX DNN 226 and BS UL RX DNN 208 for maintaining downlink or uplink transmission signals 205a 1205b that satisfy a white noise interference level whilst reducing interference to other DNN or non-DNN receivers in the cell or region around the BS 210 and UE 220. The scrambling DNN operations enable the BS 210 or UE 220 to maintain a white noise interference level when transmitting each of the downlink and/or uplink transmission signals 205a / 205b without generating accidental transmission spikes due to the multiplicity of different combinations of input communication data processed by each of the BS DL TX DNN 206 or UE UL TX DNN 226, respectively. Thus, transmission spikes in transmission signals from the BS 210 or UE 220 are mitigated, reduced and/or prevented from occurring when controlling the reconfiguration of the BS DL TX DNN 206 or UE UL TX DNN 226 and corresponding UE DL RX DNN 228 or BS UL RX DNN 208, respectively, whilst satisfying a white noise interference level. The BS DL TX DNN 206 (or UE UL TX DNN 226) and corresponding UE DL RX DNN 228 (or BS UL RX DNN 208) of the BS 210 and UE 220 (or UE 220 and BS 210), respectively, are reconfigured efficiently, rapidly, and dynamically in real-time to change the white noise interference level of downlink/uplink transmission signals 205a 1205b, while at the same time maintaining the transmission power or bit I symbol error rate of the signal of interest. A further advantage includes the efficient synchronisation between a BS 210 and UE 220 during DL I UL scrambling communications that enables dynamic whitening of the transmission signals 205a 1205b from the BS 210 or UE 220 and enables reception and decoding of the dynamically whitened transmission signal by the corresponding UE 220 or BS 210.

[0080] Although a wireless communication network I system is described with reference FIGs. 1 or 2a and/or as herein described, this is by way of example only and it is not so limited, it is to be appreciated by the skilled person that any type of communication network I system is applicable such as, for example, any telecommunication network; any wired communication network; any wireless communication network; a satellite network; a peer-2-peer communication network; a communication network using third generation (3G), fourth generation (4G), fifth generation (5G), and/or sixth generation (6G) and beyond standards technologies; a Wi-Fi communication network; optical communication network; a fibre optic communication network; and/or any other network for communications between the first device and second device; combinations thereof, modifications thereto, and/or as the application demands. Although the first device is described as BS 210 with reference to FIG. 2a and/or as herein described, this is by way of example only and it is not so limited, it is to be appreciated by the skilled person that the first device may be any type of communication device that is capable of communicating with the second device such as, without limitation, for example a UE, a BS, a satellite, a mobile phone or smart phone, a laptop, a computing device, a device using 3G, 4G, 5G, and/or 6G and beyond standards technologies, and/or any other device used for communications with the second device; combinations thereof, modifications thereto, and/or as the application demands. Although the second device is described as a UE 220 with reference to FIG. 2a and/or as herein described, this is by way of example only and it is not so limited, it is to be appreciated by the skilled person that the second device may be any type of communication device that is capable of communicating with the first device such as, without limitation, for example a UE, a BS, a satellite, a mobile phone or smart phone, a laptop, a computing device, a device using 3G, 4G, 5G, and/or 6G and beyond standards technologies, and/or any other device used for communications with the first device; combinations thereof, modifications thereto, and/or as the application demands.

[0081] FIG. 2b illustrates an example power spectral density (PSD) graph 230 representing the PSD of a transmission signal 235 satisfying a white noise interference level 232. In this example, the RF Analog transmit components of the RF front-end processes the output communication signal generated by a transmitting DNN model and upconverts the resulting transmission signal 235 to a carrier frequency of f_c with a bandwidth 231 of 2fi . The PSD of the transmission signal 235 over its entire bandwidth is below the white noise interference level 232. In this example, FIG. 2b shows the white noise interference level 232 as having a constant amplitude for a flat white noise power spectral density over the frequencies of a bandwidth of interest, i.e. , bandwidth 231 between the frequencies f_c-fi and f_c+fi . Although the white interference noise level is described as a flat white noise power spectral density, this is by way of example only, the skilled person would appreciate that the white interference noise level is any suitable measure of white noise interference or interference such as, for example, the white noise interference level may be a total power for a flat white noise power spectral density over the frequencies of a bandwidth of interest, and/or any other suitable measure of interference and the like. In this case, a spectral analysis of the transmission signal 235 indicates that the output communication signal generated by a transmitting DNN model would result in a transmission signal 235 that satisfies the white noise interference level.

[0082] FIG. 2c illustrates another PSD graph 240 representing the PSD of a transmission signal 245 that does not satisfy the white noise interference level 232. In this example, the RF Analog transmit components processes the output communication signal generated by a transmitting DNN model and upconverts the resulting transmission signal 245 to a carrier frequency of f_c with a bandwidth 231 of 2fi. The PSD of the transmission signal 245 has significant transmission spikes in the form of PSD peaks 246a, 246b, 246c, and 246d that exceed the white noise interference level 232. These PSD peaks 246a, 246b, 246c, and 246d will result in significant interference to other devices in the region of a first device 210 should it process the output communication signal for transmission. In this case, a spectral analysis of the transmission signal 245 would indicate that the output communication signal generated by a transmitting DNN model would result in a transmission signal 245 that does not satisfy the white noise interference level. As another example, even if the average PSD of the transmission signal 245 is below the white noise interference level 232, an analysis of the transmission spikes or PSD peaks 246a, 246b, 246c, and 246d of transmission signal 245 may determine the average PSD of the PSD peaks 246a, 246b, 246c, and 246d is above a tolerable transmission spike PSD threshold, and as a result, may indicate the transmission signal 245 does not satisfy the white noise interference level 232.

[0083] FIG. 2d illustrates a further example PSD graph 250 representing the PSD of a transmission signal 255 that also does not satisfy the white noise interference level 232. In this example, the RF Analog transmit components processes the output communication signal generated by a transmitting DNN model and upconverts the resulting transmission signal 255 to a carrier frequency of f_c with a bandwidth 231 of 2fi. The PSD of the transmission signal 255 over its entire bandwidth is above the white noise interference level 252. In this case, a spectral analysis of the transmission signal 255 would indicate that the output communication signal generated by a transmitting DNN model would result in a transmission signal 255 that exceeds the white noise interference level and so does not satisfy the white noise interference level.

[0084] FIG. 2e illustrates a further example PSD graph 260 representing the PSD of a transmission signal 265 that satisfies the white noise interference level 232. In this example, the RF Analog transmit components processes the output communication signal generated by a transmitting DNN model and upconverts the resulting transmission signal 265 to a carrier frequency of f_c with a bandwidth 231 of 2fi . The PSD of the transmission signal 265 has minor transmission spikes in the form of PSD peaks 266a, 266b, 266c, and 266d that exceed the white noise interference level 232. These minor PSD peaks 266a, 266b, 266c, and 266d should not result in significant interference to other devices in the region of a first device 210 should it process the output communication signal for transmission. In this case, a spectral analysis of the PSD of transmission signal 265 would indicate that the output communication signal generated by a transmitting DNN model would result in a transmission signal 265 that satisfies the white noise interference level. For example, the average PSD of the transmission signal 265 is below the white noise interference level 232, and an analysis of the minor transmission spikes or PSD peaks 266a, 266b, 266c, and 266d of transmission signal 265 may determine the average PSD of the PSD peaks 266a, 266b, 266c, and 266d is below a tolerable transmission spike PSD threshold. Given these two conditions, the analysis of the PSD of transmission signal 265 may indicate the transmission signal 265 satisfies the white noise interference level 232.

[0085] FIG. 3a illustrates an example communication system 300a including a first device 310 and a second device 320 implementing a scrambling of the input layer of the transmitting DNN structure 306 and a reciprocal descrambling of the output layer of the receiving DNN structure 308. The transmitting DNN structure 306 of the first device 310 includes a transmitting DNN model including an input neural network layer 316 and further DNN layers 307 (e.g., one or more hidden layers and an output layer) represented by the block labelled DNNi. The input neural network layer 316 receives the input communication data 301a, where the further DNN layers 307 processes the output of the input neural network layer 316 to generate output communication signal 318. The TX RF front-end components 303a processes and transmits the output communication signal 318 via antennas as transmission signal 305.

[0086] In this example, the transmitting DNN model of the transmitting DNN structure 306 is reconfigured based on a randomizing operation performed on the ordering of the neural network nodes of the input neural network layer 316. In this case, the scrambling DNN operation scrambles the nodes of the input neural network layer. Equivalent to scrambling the input neural network layer, the randomization operation can be used to scramble the input communication data 301a prior to inputting to the input neural network layer of the transmitting DNN model of the transmitting DNN structure 306. For example, the first device 310 selects the NNSI, which specifies that the input layer of the transmitting DNN model is randomized/scrambled, and selects one or more specific time slots using this configuration of the transmitting DNN structure 306. For example, the selected NNSI includes a neural network layer indicator, which specifies the input layer of the transmitting DNN model is to be randomized/scrambled for the one or more specific time slots using this configuration of the transmitting DNN structure 306. The first device 310 sends a control message to the second device 320 that includes the selected NNSI along with the specified one or more time slots in which scrambling is occurring when using the selected NNSI.

[0087] The receiving DNN structure 308 of the second device 320 includes a receiving DNN model including DNN layers 309 (e.g., an input layer and one or more hidden layers) represented by the block labelled DNN2 and an output neural network layer 317. The DNN layers 309 of the receiving DNN model receives a communication signal 319 output from the RX RF front-end components 303b after receiving the transmission signal 305 in the specific time slot. The DNN layers 309 processes the received communication signal 319 and outputs scrambled reconstructed communication data in the output neural network layer 317 of the receiving DNN structure. In this case, the receiving DNN model of the receiving DNN structure 308 has been reconfigured using the NNSI in which the output neural network layer 317 performs the inverse randomizing operations (descrambling operation) on the neural network nodes of the output neural network layer 317. The inverse randomizing operations correspond to the randomizing operations performed on the neural network nodes of the input neural network layer 316 of the transmitting DNN of the transmitting DNN structure 306. The receiving DNN structure 308 sends a descrambled reconstructed communication data 301 b that corresponds to the input communication data 301a to a data sink of the second device 320 or sends the reconstructed communication data 301 b to one or more upper protocol layers of a protocol stack of the second device 320 (e.g., to an application protocol layer of a protocol stack for use by one or more applications executing on the second device 320).

[0088] The input neural network layer 316 of the transmitting DNN structure 306 includes a set of A/ neural network nodes, where N>1 . Reconfiguring the transmitting DNN structure 306 includes performing the randomizing operation on the input neural network layer 316 of the transmitting DNN structure 306 by randomizing the set of N neural network nodes (or a subset of those N neural network nodes) of an input neural network layer 316. When the NNSI includes one or more random permutation parameters (e.g., seed, type of random function etc.), then randomizing the set of /V neural network nodes of the input neural network layer 316 may include performing a random permutation on the set of N neural network nodes of the input neural network layer 316. For example, the randomizing operation generates an /V-dimensional permutation matrix using a random permutation sequence of length N using the random permutation parameters. Randomizing the set of N neural network nodes of the input neural network layer 316 is based on randomising the ordering of the N neural network nodes by multiplying the ordering of the N neural network nodes with the /V-dimensional permutation matrix. In another example, this is equivalent to multiplying the input communication data 301a (or ordering thereof) with the /V- dimensional permutation matrix prior to input to the input neural network layer 316.

[0089] After the second device 320 has received the NNSI for the specific time slot, the second device 320 reconfigures the receiving DNN structure 308 prior to processing the received communication signal 319 corresponding to the specific time slot. Given the NNSI for this specific time slot includes a neural network layer indicator specifying the input neural network layer 316 of the transmitting DNN structure 306 is randomized/scrambled, then the receiving DNN structure 308 of the second device 320 is reconfigured by performing an inverse randomizing operation on the set of neural network nodes of the output neural network layer 317 of the receiving DNN structure 308. The output neural network layer 317 generates the descrambled reconstructed communication data 301b corresponding to the input communication data 301a. In an example, the second device 320 sends the reconstructed communication data 301 b to a data sink of the second device 320 or sends the reconstructed communication data 301b to one or more upper protocol layers of a protocol stack of the second device 320.

[0090] The output neural network layer 317 of the receiving DNN structure 308 includes a set of N neural network nodes, where N>1. When the NNSI includes random permutation parameters (e.g., seed, type of random function etc.), then a descrambling or inverse randomizing operation performs scrambling of the set of N neural network nodes of the output neural network layer 317. For example, this includes performing a random inverse permutation (or depermutation) on the set of N neural network nodes of the output neural network layer 317. For example, the descrambling or inverse randomizing operation generates an /V-dimensional permutation matrix using a random permutation sequence of length N using the random permutation parameters received in the NNSI for the specific time slot. The /V-dimensional permutation matrix is inverted to generate an inverted /V-dimensional permutation matrix. Multiplying the /V-dimensional output vector of the output neural network layer with the inverted /V-dimensional permutation matrix performs the descrambling operation, where the receiving DNN structure 308 generates descrambled reconstructed communication data 301b corresponding to the input communication data 301a.

[0091] Randomizing the input neural network layer and/or one or more hidden neural network layers provides the advantage of increasing the NNSI search space and increasing the likelihood of iteratively determining NNSI suitable for the transmitting DNN structure 306” to generate an output communication signal that when transmitted has a spectrum that satisfies the white noise interference level.

[0092] FIG. 3b illustrates an example communication system 300b including the first device 310 and the second device 320, where FIG. 3b further modifies the first device 310 and second device 320 of FIG. 3a to implement scrambling/descrambling of the output layer of the transmitting DNN structure 306’ and a reciprocal descrambling of the input layer of the receiving DNN structure 308’. In this example, the transmitting DNN structure 306’ of the first device 310 includes a transmitting DNN model including an output neural network layer 316’ and DNN layers 307’ (e.g., an input layer and one or more hidden layers) represented by DNNi. The DNN layers 307’ receives and processes the input communication data 301a, where the output neural network layer 316’ receives the DNNi processed input communication data 301a to generate output communication signal 318. The TX RF front-end components 303a processes and transmits the output communication signal 318 via antennas as transmission signal 305.

[0093] In this example, the transmitting DNN structure 306’ is reconfigured based on a scrambling or randomizing DNN operation performed on the ordering of the neural network nodes of the output neural network layer 316’ in a similar manner as described with reference to neural network nodes of the input neural network layer 316 of FIG. 3a. The receiving DNN structure 308’ of the second device 320 includes a receiving DNN model including DNN layers 309’ (e.g., one or more hidden layers and the output layer) represented by block labelled DNN2 and an input neural network layer 317’. A descrambling operation descrambles the input neural network layer 317’ as described with reference to the descrambling operation of the output neural network layer 317 of FIG. 3a. This effectively descrambles the received communication signal 319. The DNN layers 309’ processes the descrambled received communication signal 319 and outputs descrambled reconstructed communication data 301b that corresponds to the input communication data 301a.

[0094] The output neural network layer 316’ of the transmitting DNN structure 306’ includes a set of /V neural network nodes, where N>1. Reconfiguring the transmitting DNN structure 306’ includes performing the scrambling or randomizing operation on the output neural network layer 316’ of the transmitting DNN structure 306’ as described with reference to the scrambling or randomizing of the input neural network layer 316 of FIG. 3a. For example, randomizing the set of N neural network nodes of the output neural network layer 316’ is based on randomising the ordering of the N neural network nodes by multiplying the ordering of the N neural network nodes of the output neural network layer 316’ with the A/-dimensional permutation matrix. In another example, this is equivalent to multiplying the output communication signal 318 (or ordering thereof) with the /V-dimensional permutation matrix prior to input to the TX RF front-end components 303a.

[0095] After the second device 320 has received the NNSI for the specific time slot, the second device 320 reconfigures the receiving DNN structure 308’ prior to processing the received communication signal 319 corresponding to the specific time slot. Given the NNSI for this specific time slot includes a neural network layer indicator specifying the output neural network layer 316’ of the transmitting DNN structure 306’ is randomized/scrambled, then the receiving DNN structure 308’ of the second device 320 is reconfigured by performing a descrambling or inverse randomizing operation on the set of neural network nodes of the input neural network layer 317’ of the receiving DNN structure 308’ in a similar manner as described with reference to descrambling or inverse randomizing operation on the set of neural network nodes of the output neural network layer 317 of FIG. 3a. The remaining DNN layers 309’ represented by DNN2 of the receiving DNN structure 308’ generates descrambled reconstructed communication data 301b corresponding to the input communication data 301a. The second device 320 sends the reconstructed communication data 301 b to a data sink of the second device 320 or sends the reconstructed communication data 301 b to one or more upper protocol layers of a protocol stack of the second device 320. [0096] Randomizing the output neural network layer of the transmitting DNN structure 306’ provides the advantage of reducing the computational resources necessary to iteratively determine an updated NNSI because only the output neural network layer, which represents the output communication signal, needs to be processed for determining whether potential NNSI results in an output communication signal capable of conditioning a transmission signal that satisfies the white noise spectrum or white noise interference level.

[0097] FIG. 3c illustrates an example communications system 300c including the first device 310 and second device 320, where FIG. 3c further modifies the first device 310 and second device 320 of FIGs. 3a or 3b to more generally implement scrambling/descrambling of one or more hidden neural network layers 316” and 317” of the transmitting and receiving DNN structures 306” and 308”, respectively. The transmitting DNN structure 306” of the first device 310 includes a transmitting DNN model including one or more hidden neural network layers 316” for use in a hidden layer scrambling operation, and further DNN layers 307” (e.g., an input layer, one or more hidden layers if any, and an output layer) represented by the block labelled DNNi. The DNN layers 307” receive and process the input communication data 301a and pass the processed data to the one or more hidden neural network layers 316” for scrambling accordingly, which passes the scrambled processed data to the output neural network layer of the DNNi for generating the output communication signal 318. The TX RF front-end components 303a processes the output communication signal 318 for transmission via antennas as transmission signal 305.

[0098] In this example, the transmitting DNN structure 306” is reconfigured based on a scrambling DNN operation or randomizing operation performed on the ordering of the neural network nodes of the one or more hidden neural network layers 316” in a similar manner as described with reference to the neural network nodes of the input neural network layer 316 or output neural network layer 316’ of FIGs. 3a or 3b, respectively.

[0099] The receiving DNN structure 308” of the second device 320 includes a receiving DNN model including DNN layers 309” (e.g., an input neural network layer, one or more hidden layers if any, and an output neural network layer) represented by DNN2 and one or more hidden neural network layers 317” for descrambling, i.e., which perform the descrambling operation. The input neural network layer of the DNN2 309” receives a communication signal 319 output from the RX RF front-end components 303b after reception of the transmission signal 305 in the specific time slot. In this case, the receiving DNN structure 308” has been reconfigured using the NNSI in which the descrambling one or more hidden neural network layers 317” perform the inverse randomizing operations (descrambling operation) on the neural network nodes of the corresponding hidden neural network layers 317” in a similar manner as described with reference to the neural network nodes of the output neural network layer 317 or input neural network layer 317’ of FIGs. 3a or 3b, respectively. This effectively descrambles the received communication signal 319, where the hidden neural network layers 317” and DNN layers 309” process the communication signal 319 and outputs descrambled reconstructed communication data 301 b that corresponds to the input communication data 301a. Symmetric DNN architectures for transmitting and receiving DNN structures 306” and 308” simplify scrambling and descrambling operations.

[00100] Randomizing the one or more hidden neural network layers provides the advantage of increasing the NNSI search space and increasing the likelihood of iteratively determining NNSI suitable for the transmitting DNN structure 306” to generate an output communication signal that when transmitted has a spectrum that satisfies the white noise interference level.

[00101] Referring to FIGs. 3a to 3c, although the scrambling of the input layer, output layer and an l-th hidden layer or (L-l+1)-th hidden layer of the transmitting DNN structure 306 were described separately, this is by way of example only, it is to be appreciated by the skilled person that one or more of the input layer, output layer and hidden layers of the transmitting DNN structure 306 of the first device 310 may be scrambled and/or randomized. More generally, the NNSI specifies which neural network layers are scrambled along with scrambling parameters, where a third generation partnership project (3GPP) standard and the like may predefine the neural network layers to be scrambled. Alternatively or additionally, the NNSI specifies a neural network layer indicator (e.g., flag, bit, or field) specifying one or more neural network layers of the transmitting DNN structure 306 for scrambling. More generally, the NNSI includes a neural network layer indicator specifying scrambling of one or more neural network layers of the transmitting DNN structure 306. The transmitting DNN structure 306 includes at least one input layer, one or more hidden layers, and an output layer. The first device 310 uses the NNSI scram ble/randomize the one or more neural network layers indicated. For each l-th neural network layer of the one or more neural network layers, for 1<I<L and L is the number of neural network layers in which 1=1 is the input layer and l=L is the output layer, the l-th neural network layer of the transmitting DNN structure 306 is randomized/scrambled, whilst the second device 320 uses the corresponding NNSI (received in a control message for a specific time slot) to reconfigure the receiving DNN structure 308 by performing an inverse randomizing operation on the set of neural network nodes of the (L-l+1)-th neural network layer of the receiving DNN.

[00102] When the NNSI includes random permutation parameters, the l-th neural network layer of the transmitting DNN structure 306 is reconfigured by: generating an A/-dimensional permutation matrix based on the random permutation for the l-th neural network layer, A/> 1 and N is the number of neural network nodes in the l-th neural network layer, and multiplying the specific ordering of the set of /V-neural network nodes of the l-th neural network layer with the /V-dimensional permutation matrix to form a randomized ordering of the set of N neural network nodes. The first device 310 uses the reconfigured transmitting DNN structure 306 to process the input communication data 301a and generate a scrambled output communication signal 318. The TX RF front-end components 303a processes and transmits the output communication signal 318 via antennas as a transmission signal 305 to the second device 320 in the specific time slot. The transmission signal 305 satisfies the white noise interference level because the NNSI was selected to meet this criterion.

[00103] The second device 320 receives the corresponding NNSI including random permutation parameters, the specific one or more neural network layers, and the specific time slot(s) or scrambling timing information associated with when the receiving DNN structure 308 should be reconfigured for descrambling the corresponding received communication signal 319. At the appropriate time, for example, prior to processing the received communication signal 319 corresponding to the specific time slot, each of the specified one or more neural network layers of the receiving DNN structure 308 are reconfigured for descrambling based on the corresponding NNSI. The (L-l+1)-th neural network layer of the receiving DNN structure 308 has a set of N neural network nodes, N>1 , with a specific ordering is reconfigured by: generating a /V-dimensional permutation matrix based on the random permutation for the (L-l+1)-th neural network layer, inverting the N- dimensional permutation matrix to generate an inverted /V-dimensional permutation matrix, and multiplying the specific ordering of the set of A/-neural network nodes of the (L-l+1)-th neural network layer with the inverted /V-dimensional permutation matrix to form an inverse randomized ordering of the set of N neural network nodes. After reconfiguration, the receiving DNN structure 308 processes the received communication signal(s) 319 received in the specific time slot(s) to generate reconstructed communication data 301 b, which corresponds to the input communication data 301a transmitted in the specific time slot(s).

[00104] FIG. 4a illustrates an example transmitting DNN scrambling process 400 for generating a scrambled output communication signal from a transmitting DNN model that satisfies a white noise interference level when transmitted. Reference numerals from FIG. 1 are reused for similar or the same components, features and/or devices in FIGs. 4a to 4d. In this example, a first device 104a is in communication with a second device 104b, where the first device 104a includes a transmitting DNN structure 106 with a transmitting DNN model 107 and the second device includes a receiving DNN structure 108 with a receiving DNN model 109. The transmitting DNN model 107 has been trained to replace several functions (e.g., encoding, interleaving, scrambling, precoding, etc.) of a transmit processing chain. The receiving DNN model 109 has also been trained to perform the reciprocal operations of the transmitting DNN model 107 for generating reconstructed communication data 101b’. The transmitting DNN scrambling process 400 includes the following steps of:

[00105] In step 402, retrieving, by the first device 104a, input communication data 101a’ for transmission in a specific time slot. The input communication data 101a’ may be a bit stream or any type of digital data from a data source.

[00106] In step 404, processing, by the first device 104a, the retrieved input communication data 101a’ with the transmitting DNN structure 106 for generating an output communication signal 118 for transmission to the second device 104b.

[00107] In step 406, checking, by the first device 104a, whether transmission of the generated output communication signal 118 would satisfy a white noise interference level. If transmission of the generated output communication signal 118 would satisfy the white noise interference level (e.g., ‘Y’), then proceed to step 412, otherwise (e.g., ‘N’) proceed to step 408. The first device 104a performs a spectral analysis to determine whether the output communication signal 118 would result in a transmission signal 105a’ that satisfies the white noise interference level. That is, first device 104a uses the spectral analysis to forecast whether the output communication signal 118 will result in a transmission that satisfies the white noise interference level. For example, the spectral analysis includes calculating the power spectral density of a forecast transmission signal that would result from transmitting the output communication signal 118 after RF Analog TX processing. The forecasted transmission signal is not transmitted, rather it is an estimate of the transmission signal that would result should the output communication signal 118 be transmitted by the first device 104a.

[00108] In step 408, performing, by the first device 104a, a scrambling DNN operation in response to a transmission of the generated output communication signal not satisfying a white noise interference level by selecting NNSI for reconfiguring the transmitting DNN model 107 of the transmitting DNN structure 106 to process the input communication data 101a’ to generate a scrambled output communication signal, which when transmitted, satisfies the white noise interference level.

[00109] For example, the NNSI may include data representative of randomizing the orders of the inputs, outputs, or a set of neural network nodes of one or more specified neural network layers of the transmitting DNN model. As an example, the NNSI may include one or more random permutation parameters for randomizing the ordering of the set of neural network nodes in the specified one or more neural network layers of the transmitting DNN model 107 of the first device 104a. Randomizing the ordering uses the random permutation parameters to perform a random permutation of the ordering of the set of neural network nodes in the specified one or more neural network layers of the transmitting DNN model 107.

[00110] In step 410, reconfiguring, by the first device 104a, the transmitting DNN model 107 based on the selected NNSI. Proceed to step 404, where the reconfigured transmitting DNN model 107 processes the input communication data 101a’ retrieved in step 402. For example, reconfiguration of the transmitting DNN model includes reconfiguring the transmitting DNN model 107 using the NNSI to randomize an ordering of the inputs, outputs, or a set of neural network nodes of one of more specific neural network layers as specified by the NNSI.

[00111] In step 412, checking, by the first device 104a, whether the NNSI has been updated. If the NNSI has been updated (e.g., ‘Y’), i-^e-, the updated NNSI is different to the previously selected NNSI, then proceed to step 414, otherwise (e.g., ‘N’), proceed to step 416.

[00112] In step 414, transmitting, from the first device 104a to the second device 104b, a control message including data representative of an indication of the NNSI and scrambling timing information including the specific time slot (e.g., one or more time slots) for directing when the second device 104a is to reconfigure the receiving DNN model 109 for generating reconstructed communication data 101b’ corresponding to the input communication data 101a’. Proceed to step 416.

[00113] In step 416, transmitting, by the first device 104a, the scrambled output communication signal 118 to the second device 104b with respect to the scrambling timing information, where transmission of the scrambled output communication signal satisfies the white noise interference level. Proceed to step 402 for retrieving further input communication data 101a’ for transmission.

[00114] FIG. 4b illustrates an example spectral analysis process of step 406 of FIG. 4a for analysing whether the spectral density of a transmission signal representing the output communication signal generated by the transmitting DNN model 107 satisfies the white noise interference level. The spectral analysis process includes the following steps of:

[00115] In step 406a, estimating the spectral density of a forecast transmission signal representing the output communication signal 118 after processing for transmission. The transmission signal is a forecast transmission signal because it has not yet been transmitted. For example, the spectral density of the forecast transmission signal is estimated by simulating or modelling the characteristics of the RF front-end TX components, antennas, and communication channel given the output communication signal 118. Additionally or alternatively, the estimated spectral density of the forecast transmission is estimated for each antenna output of the RF front-end TX components and the like. The estimated spectral density of the forecast transmission signal can be estimated based on combining estimates of the spectral density of the forecast transmission at the output of each of the antennas of the RF front-end TX components.

[00116] In step 406b, checking whether the estimated spectral density of the forecast transmission signal satisfies the white noise interference level. If the estimated spectral density of the forecast transmission signal satisfies the white interference noise level (e.g., ‘Y’), then proceed to step 406d, otherwise (e.g., ‘N’) proceed to step 406c.

[00117] The checking further includes detecting or identifying whether the estimated spectral density satisfies the white noise interference level. For example, an analysis of the power spectral density of the transmission signal detects or identifies that the spectral density of the forecast transmission signal does not satisfy the white noise interference level when it forms one or more spikes of interference above the white noise interference level. In another example, an analysis of the power spectral density of the forecast transmission signal detects or identifies that the spectral density of the forecast transmission signal does satisfy the white noise interference level when the power spectral density of the transmission signal over the bandwidth of interest is below the white noise interference level or within a predetermined threshold region of the white noise interference level.

[00118] In a further example, the checking and analysis of 406b includes comparing the estimated spectral density of the forecast transmitted signal with a white noise spectral density associated with the white noise interference level. In response to the estimated spectral density of the forecast transmitted signal being less then or substantially matching the white noise spectral density associated with the white noise interference level, then proceed to step 406d. Otherwise, in response to the estimated spectral density of the forecast transmitted signal being greater than or substantially diverging from the white noise spectral density associated with the white noise interference level, then proceeding to step 406c.

[00119] In step 406c, indicating data representative of an indication the white noise interference level is not satisfied. The indication may be based on setting a predetermined negative flag/field value (e.g., ‘O’, ‘N’, negative binary value) that is indicative of the white interference level being satisfied.

[00120] In step 406d, indicating data representative of an indication the white noise interference level is satisfied. The indication may be based on setting a predetermined positive flag/field value (e.g., ‘1’, ‘Y’, positive binary value) that is indicative of the white interference level being satisfied.

[00121] For example, a processor or other computing device (e.g., BS DNNC 214 or UE DNNC 224 of FIG. 2) automatically performs the steps of spectral analysis process of step 406 of FIG. 4a. Although several examples of performing the check in step 406b in relation to whether the forecast transmission signal satisfies a white interference noise level have been described, this is by way of example only, it is to be appreciated by the skilled person that other suitable automatic methods or techniques are applicable for checking whether the forecast transmission signal satisfies a white interference noise level and the like.

[00122] FIG. 4c illustrates an example control message process 430 at a second device 104b for receiving one or more control messages including NNSI transmitted from the first device 104a in step 414 of FIG. 4a. The control message process 430 includes the following steps of:

[00123] In step 432, receiving, at the second device 104b from the first device 104a, a control message including NNSI and scrambling timing information including a specific time slot associated with the NNSI. The specific time slot indicates when the first device 104a will transmit a transmission signal 105a’ representing output communication signal generated from a transmitting DNN model 107 during a scrambling DNN operation.

[00124] In step 434, storing, by the second device 104b, the received NNSI mapped with the specific time slot (or scrambling timing information). The second device 104b retrieves the stored NNSI mapped with the specific time slot for reconfiguring the receiving DNN model 109 prior to the second device 104b processing a received communication data signal corresponding to reception of a transmitted signal 105a’ in the specific time slot. [00125] FIG. 4d illustrates an example descrambling DNN process 440 for receiving, at a second device 104b, a transmission of output communication signal from a transmitting DNN model 107 of a first device 104a and generating reconstructed communication data 101b’. The descrambling DNN process 440 includes the following steps of:

[00126] In step 442, the second device 104b receives a communication data signal 119 based on reception of a transmitted signal 105a’ transmitted from the first device according to scrambling timing information including a specific time slot. For example, the received communication data signal 119 is based on receiving a transmit signal 105a’ that results from transmitting the output communication signal generated in transmitting DNN scrambling process 400 of FIG. 4a, where the transmission of the output communication signal occurs in the specific time slot or according to specific scrambling timing information.

[00127] In step 444, the second device 104b retrieves NNSI, if any, for the specific time slot of the scrambling timing information. As described in FIG. 4c, the NNSI is received from one or more control messages sent by the first device 104a and stored at the second device 104b. The second device 104b maps the received NNSI to the corresponding specific time slot(s) that the NNSI is to be used for.

[00128] In step 446, the second device 104b checks whether any NNSI has been retrieved in relation to the specific time slot. For example, the second device 104b checks whether the NNSI has changed from a previous NNSI for the specific time slot. If NNSI has not been retrieved or has not changed (e.g., ‘N’), then proceed to step 450, where the current receiving DNN model 109 is used for descrambling the received communication data signal 119. If NNSI has been retrieved or has changed (e.g., ‘Y’), then proceed to step 448, where the receiving DNN model 109 is reconfigured according to the NNSI.

[00129] In step 448, the second device 104b configures (if the first time) or reconfigures the receiving DNN model 109 of the second device 104b according to the NNSI for processing the received communication data signal 119 according to the scrambling timing information or the specific time slot. Proceed to step 450. [00130] For example, the NNSI includes data representative of randomizing the orders of the inputs, outputs, or a set of neural network nodes of one or more specified neural network layers of the transmitting DNN model 107 of the first device 104a. The reconfiguring of the receiving DNN model 109 of the second device 104b further includes reconfiguring the receiving DNN model 109 of the second device 104b using the NNSI to reverse the randomisation applied to the specified neural network layers of the transmitting DNN model 107.

[00131] In step 450, the second device 104b processes the received communication data signal 119 with the receiving DNN model 109 for reconstructing communication data represented by the received communication data signal 119.

[00132] In step 452, the second device 104b sends the reconstructed communication data 101 b’ to a data sink and/or one or more upper layers of a protocol stack of the second device 104b.

[00133] FIG. 5 illustrates an example DNN communication system 500 with a first device 510 in communication with a second device 520. In this example, the first device 510 includes a transmitting DNN structure 506 (or TX DNN 506), a TX DNN Controller 514, an output communication (OC) signal buffer 536a (also referred to as OC buffer 536a), an NNSI buffer 537a, and an RF front-end TX componentry 503a. Input communication data 501a that is to be transmitted in each specific time slot (e.g., time slot q (TS(q)), TS(q+1), , TS(q+Q) and so on) is applied to the input of the TX DNN 506. The TX DNN 506 generates corresponding OC signals 518a-518q for transmission in each of the specific time slots (e.g., TS(q), TS(q+1), , TS(q+Q) and so on). Each of the generated OC signals 518a to 518q passes through the TX DNN Controller 514, which operates as a gate for determining whether or not an OC signal 518a should be passed to RF front-end TX componentry 503a for transmitting the OC signal 518a as a transmission signal 505a in TS(q) that satisfies a white noise interference level. The TX DNN Controller 514 implements the scrambling DNN operations performed at the first device 510 as described with reference to FIGs. 1a to 4b and/or as described herein. If the TX DNN Controller 514 detects that the generated OC signal 518a will satisfy the white noise interference level when the first device 104a transmits it in TS(q) as a transmission signal 505a, then the TX DNN Controller 514 sends the OC signal 518a to the OC buffer 536a for transmission in the specific time slot TS(q). If the TX DNN Controller 514 detects that the generated OC signal 518a will not satisfy the white noise interference level should the first device 504a transmit it as transmission signal 505a in specific TS(q), then the TX DNN Controller 514 does not allow the generated OC signal 518a into the OC buffer 536a. Instead, the TX DNN Controller 514 selects an NNSI for the specific time slot TS(q) for reconfiguring the TX DNN 506 to generate an OC signal 518a for the specific time slot TS(q) that is forecast to satisfy the white noise interference level when transmitted in the specific time slot TS(q).

[00134] For example, the TX DNN Controller 514 selects an NNSI signal 527a from an NNSI table (not shown) in storage, each NNSI entry is associated with a particular white noise interference level characteristic, which is likely to result in the resulting OC signal 518a to satisfy the white noise interference level when transmitted. The TX DNN Controller 514 directs the TX DNN 506 to be reconfigured for a scrambling DNN operation based on the selected NNSI signal 527a as described with reference to FIGs. 1a to 4b and/or as herein described, where the reconfigured TX DNN 506 reprocesses the input communication data 501a for the specific time slot TS(q) to generate an updated or rescrambled OC signal 518a. If the TX DNN Controller 514 detects that the generated OC signal 518a will still not satisfy the white noise interference level should the first device 510 transmit it as transmission signal 505a, then the TX DNN Controller 514 iteratively performs selection of new NNSI signal until the OC signal 518a satisfies the white noise interference level when transmitted. If the TX DNN Controller 514 detects that the generated OC signal 518a will now satisfy the white noise interference level should the first device 510 transmit it as a transmission signal 505a, then the TX DNN Controller 514 sends the OC signal 518a to the OC buffer 536a for transmission in the specific time slot TS(q). The TX DNN Controller 514 also sends the corresponding selected NNSI signal 527a to the NNSI buffer 537a. Each NNSI signal 527a includes data representative of the specific time slots that it is applicable for (e.g., TS(q)), and an indication of the randomization/scrambling parameters used to reconfigure the TX DNN 506, which an RX DNN Controller 524 may use for reconfiguring the RX DNN 508 for descrambling and generating reconstructed communication data 501 b. The first device 510 transmits each NNSI signal 527a in the NNSI buffer 537a in a control message 505b to the second device 520 at the appropriate time (e.g., prior to transmission of the corresponding OC signal 518a) for directing when the second device 520 should reconfigure the RX DNN 508 for processing a corresponding received OC signal for the specific time slot TS(q) and descrambling to generate reconstructed communication data 501 b for the specific time slot TS(q). Typically, the first device 510 sends the selected NNSI signal 527a for the specific time slot TS(q) prior to transmission of the OC signal 518a in the specific time slot TS(q). In other examples, the first device 510 transmits the selected NNSI signal 527a in a control message after transmission of the OC signal 518a, where the second device 520 uses the received NNSI signal 527a for reconfiguring the RX DNN 508 prior to processing the corresponding received OC signal.

[00135] Each of the OC signals 518a-518q and/or NNSI signals 527a-527p that are in the OC buffer 536a and NNSI buffer 537a are output for RF processing at the appropriate time. For example, each of the OC signals 518a-518q will be RF processed for transmission in their specific time slots TS(q), TS(q+1), , TS(q+Q) by the RF front-end TX componentry 503a and transmitted a data transmission signal 505a over a data communication channel. For example, each of the NNSI signals 527a-527p will be RF processed for transmission prior to the specific time slots TS(q), TS(q+1), , TS(q+Q) used for transmission of the OC signals 518a-518q by the RF front-end TX componentry 503a and transmitted as a control message 505b over a control communications channel. In another example, each of the NNSI signals 527a-527p will be RF processed for transmission prior to the second device 520 processing the corresponding received OC signals 518a-518q by the RF frontend TX componentry 503a and transmitted as one or more control messages 505b over the control communication channel.

[00136] For example, the data communication channel includes a downlink data channel, for example, a physical downlink shared channel (PDSCH), when the first device 510 is a base station and the second device 520 is a user equipment. Alternatively, the data communication channel includes an uplink data channel, for example, a physical uplink shared channel (PLISCH), when the first device 510 is a user equipment and the second device 520 is a base station. Similarly, the control communication channel includes a downlink control channel, for example, a physical downlink control channel (PDCCH), when the first device 510 is a base station and the second device 520 is a user equipment. Alternatively, the control communication channel includes an uplink control channel, for example, a physical uplink control channel (PUCCH), when the first device 510 is a user equipment and the second device 520 is a base station.

[00137] The second device 520 includes a receiving DNN structure 508 (or RX DNN 508), a RX DNN Controller 524, a received output communication (Rx OC) signal buffer 536b (also referred to as Rx OC buffer 536b), an NNSI buffer 537b, and a RF front-end RX componentry 503b. The RF front-end RX componentry 503b receives the transmission signal 505a over the data communication channel in a specific time slot (e.g., TS(Q+1 )), processes the received transmission signal 505a to generate a baseband received OC data signal (Rx OC signal) 519a. The RF front-end RX componentry 503b feeds the Rx OC signal 519a into the Rx OC buffer 536b for the specific time slot (e.g., TS(Q+1)). In this example, the Rx OC buffer 536b is a first-in first-out (FIFO) queue, but can be any type of buffer/queue implementation. From time-to-time, the RF front-end RX componentry 503b also receives a control message 505b including NNSI signal 529a for a specific time slot (e.g., TS(Q+1)) transmitted over the control communication channel, processes the received control message 505b to generate a NNSI signal 529a. The RF front-end RX componentry 503b feeds the NNSI signal 529a into the NNSI buffer 537b for the specific time slot (e.g., TS(Q+1)). The Rx OC buffer 536b connects to the RX DNN Controller 524, which receives from Rx OC buffer 536b an Rx OC signal 519q corresponding to a specific time slot (e.g., TS(Q+P)). The RX DNN Controller 524 implements the reciprocal descrambling DNN operations to the scrambling DNN operations that were performed at the first device 510 as described with reference to FIGs. 1a to 4d and/or as described herein.

[00138] For example, the RX DNN Controller 524 controls when to input the Rx OC signal 519q for the specific time slot TS(Q+P) into the RX DNN 508 for generating reconstructed communication data 501 b corresponding to the input communication data 501 a for the specific time slot TS(Q+P). In this case, the Rx DNN Controller 524 identifies there is an NNSI signal 529q for the specific time slot TS(Q+P) for reconfiguring the RX DNN 508 to descramble the Rx OC signal 519q received in time slot TS(Q+P). Prior to processing the Rx OC signal 519q for time slot TS(Q+P), the RX DNN Controller 524 reconfigures the RX DNN 508 using the NNSI signal 529q for time slot TS(Q+P). After reconfiguration, the RX DNN Controller 524 inputs the Rx OC signal 519q into the reconfigured RX DNN 508 for generating the reconstructed communication data 501 b for the specific time slot TS(Q+P). The second device 520 sends the reconstructed communication data 501 b for the specific time slot TS(Q+P) to a data sink of the second device 520 or sends the reconstructed communication data 501b to one or more upper protocol layers of a protocol stack of the second device 520.

[00139] As can be seen, the second device 520 receives several NNSI signals 529a, 529b, and 529p in the NNSI buffer 537b and which correspond to Rx OC signals 519a, 519b, and 519p for specific time slots TS(Q+1), TS(Q+2), and TS(Q+P-1). The RX DNN Controller 524 reconfigures the RX DNN 508 using the corresponding NNSI signals 529a, 529b, 529o, and 529p prior to processing each of the corresponding Rx OC signals 519a, 519b, and 519p. The TX DNN Controller 514 might not necessarily update the NNSI signal for each time slot because the TX DNN Controller 514 may find the same NNSI is applicable for multiple consecutive time slots, in which case, the RX DNN Controller 524 only needs to reconfigure the RX DNN 508 when a control message is received with an updated NNSI for some subsequent time slots. For example, there may be no further NNSI changes between TS(Q+P-2) and TS(Q+2), thus the RX DNN Controller 524 only needs to reconfigure the RX DNN 508 using NNSI signal 529o until after the Rx OC signal 519b is to be processed at TS(Q+2), so the Rx OC signals 519c up to but not including Rx OC signal 519p are processed by the same configuration of RX DNN 508 for time slots TS(Q+3) to TS(Q+P-2).

[00140] FIG. 6a illustrates example permutation operations 600 for performing a random permutation on a specific ordering of one or more neural network nodes 641 of a neural network layer of a transmitting DNN model for scrambling DNN operation, and a random inverse permutation (also referred to as a depermutation) on a specific ordering of a set of neural network nodes 642 of a neural network layer of a receiving DNN model for descrambling DNN operations.

[00141] In this example, a set of neural network nodes 641 of a particular neural network layer of the transmitting DNN model have a specific ordering indicated, for example, by node labels 1 , 2, 3, 4, and 5. The transmitting DNN model is reconfigured by randomizing the specific ordering of the set of neural network nodes 641 for the particular neural network layer using a random permutation operation 616. The random permutation operation 616 permutes or randomises the specific ordering of the neural network nodes 641 resulting in a permuted or randomized ordering of neural network nodes 642 for that particular neural network layer of the transmitting DNN model. The specific ordering of the set of neural network nodes

642 for the particular neural network layer after the random permutation operation 616 is performed is indicated, for example, by the reordered node labels 2, 1 , 3, 5, 4. The input connections to neural network node 1 before the random permutation operation 616 are now input to neural network node 2, the input connections to neural network node 2 before the random permutation operation 616 are now input to neural network node 1 , the input connections to neural network node 3 before the random permutation operation 616 are still input to neural network node 3, the input connections to neural network node 4 before the random permutation operation 616 are now input to neural network node 5, and the input connections to neural network node 5 before the random permutation operation 616 are now input to neural network node 4. Even though the outputs of the set of neural network nodes 642 (e.g., nodes 1 , 2, 3, 4 and 5) are still connected to the same neural network nodes in other neural network layers of the reconfigured transmitting DNN model, the random permutation of the inputs to the set of neural network nodes 642 results in a scrambling of the resulting transmitting DNN model output. This may be referred to as a scrambling DNN operation.

[00142] For the particular random permutation operation 616, NNSI includes the necessary data to reverse the random permutation operation 616 for use by a second device in reconfiguring a corresponding receiving DNN model for reversing the random permutation operation 616. The second device uses the NNSI to identify the reciprocal neural network layer of the receiving DNN model and reconfigured the receiving DNN model using a random inverse permutation operation or a random depermutation operation 617 on the specific ordering of the set of neural network nodes of the identified neural network layer of the receiving DNN model. In this example, a set of neural network nodes 643 of a particular neural network layer of the receiving DNN model have a specific ordering indicated, for example, by node labels 2, 1 , 3, 5, 4. The receiving DNN model is reconfigured by randomizing the specific ordering of the set of neural network nodes 643 for the particular neural network layer using a random depermutation operation 617. The random depermutation operation 617 performs a depermutation on the specific ordering of the neural network nodes

643 resulting in a depermuted or derandomized ordering of neural network nodes

644 for that particular neural network layer of the receiving DNN model. The specific ordering of the set of neural network nodes 644 for the particular neural network layer after the random depermutation operation 617 is performed is indicated by the reordered node labels 1 , 2, 3, 4, 5. This means that the input connections to neural network node 1 before the random depermutation operation 617 are now input to neural network node 2, the input connections to neural network node 2 before the random depermutation operation 617 are now input to neural network node 1 , the input connections to neural network node 3 before the random depermutation operation 617 are still input to neural network node 3, the input connections to neural network node 4 before the random depermutation operation 617 are now input to neural network node 5, and the input connections to neural network node 5 before the random depermutation operation 617 are now input to neural network node 4. Even though the outputs of the neural network nodes 1 , 2, 3, 4 and 5 are still connected to the same neural network nodes in other neural network layers of the reconfigured receiving DNN model, the random permutation of the inputs to the set of neural network nodes 644 results in a descrambling of the resulting receiving DNN model output. This may be referred to as a descrambling DNN operation.

[00143] The NNSI may include, for example, randomisation parameters /functions for randomizing the orders of the inputs, outputs, or a set of neural network nodes of one or more specified neural network layers of the transmitting DNN that are to be randomized. The one or more specific neural network layers of the transmitting DNN include, without limitation, for example a selection of one or more of: the input neural network layer (or input layer), the output neural network layer (or output layer), and/or a hidden neural network layer, and/or a combination thereof, of the transmitting DNN model. The specified neural network layers of the transmitting DNN model are randomised/scrambled. The NNSI is used for reconfiguring the transmitting and receiving DNN models for performing scrambling DNN operations.

[00144] The second device uses the NNSI to reconfigure the receiving DNN model in synchronisation with the reception of the transmitted signal representing the output communication signal of the reconfigured transmitting DNN model. For performing random permutation operations, the NNSI further includes data representative of one or more random permutation parameters and/or functions including one or more of: seed data, a random permutation iteration or order number, or identification of a seed generation and pseudo-randomizing function for performing the random permutation of the specific ordering of a set of neural network nodes of a specified neural network layer of a transmitting DNN model. The particular seed data may include, without limitation, for example the identity of second device (e.g., UE), identity of first device (e.g., BS)), an identity of a cell the second device is located, an identity of the cell the first device is located, timing slot information, frame identification number, and/or any other information associated with the first or second device, a particular random permutation iteration or order number. For example, an initial seed or initial seed value may be generated from the seed data using the seed generation and pseudorandomizing function for use in randomizing and/or derandomizing the neural network nodes of the specified one or more neural network layers of the transmitting DNN model (TX DNN) and/or receiving DNN model (RX DNN), respectively.

[00145] As an example, the second device receives NNSI including one or more random permutation parameters such as, for example, seed data, a random permutation iteration or order number, or identification of a seed generation, indication of the one or more neural network layers that were scrambled by at the first device, and pseudo-randomizing function for use in performing the random permutation of the set of neural network nodes of the indicated one or more neural network layers. The second device also receives corresponding timing information associated with when to reconfigure the RX DNN (e.g., descramble corresponding neural network layers of the RX DNN). The timing information including one or more time slots. At a particular time slot, reconfiguring the RX DNN of the second device may include generating, for each neural network layer of the RX DNN corresponding to each indicated neural network layer of the TX DNN, an inverse random permutation sequence corresponding to the random permutation iteration or order number. The inverse random permutation sequence has a length equal to the number of neural network nodes in the set of neural network nodes of said each neural network layer of the RX DNN. Derandomizing or performing an inverse permutation, for each neural network layer of the RX DNN corresponding to each neural network layer of the TX DNN, the set of neural network nodes of said each neural network layer of the RX DNN by applying the generated inverse random permutation sequence to the set of neural network nodes. Alternatively, an inverse random permutation matrix can be generated and applied to the ordering of the set of neural network nodes and the like.

[00146] FIG. 6b illustrates an example random permutation sequence 650 starting from an initial seed 652 (also referred to as an initial seed value 652). For a particular neural network layer with a set of A/ neural network nodes, the initial seed 652 is input to a random permutation function for generating a first random permutation sequence 651a of size /V. The first random permutation sequence 651a may be a permutation of integers in the range [1 , /V], where the integers represent node labels of the set of N neural network nodes. The first random permutation sequence 651a results from permutation iteration 1 of the random permutation function with initial seed 652. Further iterations or cycles of the random permutation function with initial seed 652 result in different random permutation sequences of size N such as, for example, a second random permutation sequence 651 b in a second permutation iteration 2, , a third random permutation sequence 651 c in a third permutation iteration 3, , an i-th random permutation sequence 651 i in a i-th permutation iteration i, , an n-th random permutation sequence 651 n in an n-th permutation iteration n, and so on. When given an initial seed value 652 and random permutation function pair, then an i-th random permutation sequence of length N (or i-th cycle of a random permutation sequence of length A/) can be generated by simply specifying the i-th random permutation iteration number and cycling through the random permutation sequences generated using the initial seed 652 and random permutation function until the i-th cycle.

[00147] As an example, an initial control message may include NNSI with data representative of an initial seed value, the random permutation function indication, a random permutation iteration number, and a specified one or more neural network layers of the transmitting DNN model that have been randomised as described with reference to FIGs. 1a to 6a. The first device transmits the initial control message to the second device for reconfiguring the corresponding neural network layers of the receiving DNN model as described with reference to FIGs. 3a to 3c using the received initial seed value, random permutation function indication, random permutation iteration number and the specified neural network layer, for generating the corresponding random permutation sequence associated with the random permutation iteration number. The generated random permutation sequence generates the random depermutation (or inverse permutation) operation 617 for use in reconfiguring the corresponding neural network layer of receiving DNN model at the second device for the descrambling DNN operation and generation of reconstructed communication data. The second device sends the reconstructed communication data to a data sink of the second device or sends the reconstructed communication data to one or more upper protocol layers of a protocol stack of the second device. Furthermore, the first device sends subsequent updates of NNSI used to reconfigure the transmitting DNN model in subsequent control messages, which include an indication of the updated random permutation iteration number. When the initial seed and/or random permutation function are updated or changed, then these are sent in a further control message or one or more further control messages.

[00148] FIG. 6c illustrates permutation and depermutation (or inverse permutation) operations 660 in which the random permutation operation 616 uses a random permutation matrix 616’ resulting from a selected i-th random permutation sequence 652i of FIG. 6b for use in scrambling I descrambling one or more neural network layers of a transmitting/receiving DNN model. The i-th random permutation sequence 652i is used to generate an i-th N x N random permutation matrix Pi 616’. For example, calculating the i-th N x N random permutation matrix Pi 616’ by: generating an Nx N identity matrix I, where each of the columns are consecutively labelled from 1 to N; and then permuting the columns of the identify matrix I using the generated /- th random permutation sequence 652i. This results in the random permutation matrix Pi 616’, which represents the i-th random permutation sequence 652i. The random permutation operation 616 is performed by multiplying the random permutation matrix Pi 616’ with an /V-dimensional vector resulting in a permuted A/-dimensional vector that is permuted according to the i-th random permutation sequence 652i. After the random permutation matrix Pi 616’ has been calculated, the random depermutation operation 617 is derived based on taking the matrix inverse of permutation matrix Pi 616’ resulting in a depermutation matrix Di 617’ where Di = (Pi)'¹. The random depermutation operation 617 may be performed by multiplying the depermutation matrix Di 617’ with an /V-dimensional vector requiring depermutation resulting in a depermuted /V-dimensional vector.

[00149] For example, for a selected NNSI indicating an i-th random permutation iteration number of the i-th random permutation sequence 652i, the l-th neural network layer of a transmitting DNN model, where 1<I<L and 1=1 is the input layer, l=L is the output layer and L is the number of neural network layers including one or more hidden neural network layers, is reconfigured by: generating the i-th random permutation sequence 652i, generating an /V-dimensional random permutation matrix Pi 616’ using the i-th random permutation sequence 652i for the I-th neural network layer, /V> 1 and N is the number of neural network nodes in the I-th neural network layer, and multiplying a specific ordering vector representing the current specific ordering of the set of /V-neural network nodes of the I-th neural network layer with the /V-dimensional random permutation matrix Pi 616’ resulting in a randomized ordering vector representing a permuted set of N neural network nodes.

[00150] For example, the first device selects NNSI indicating an i-th random permutation iteration number of the i-th random permutation sequence 652i and indicating that the I-th neural network layer of the transmitting DNN model is reconfigured. The first device sends the selected NNSI to the second device in a control message as described herein. Given the selected NNSI, the second device reconfigures the (L-l+1)-th neural network layer of the receiving DNN model by: generating the i-th random permutation sequence 652i, generating an /V-dimensional random permutation matrix Pi 616’ using the i-th random permutation sequence 652i for the (L-l+1)-th neural network layer, /> 1 and N is the number of neural network nodes in the (L-l+1)-th neural network layer, generating an /V-dimensional depermutation matrix Di 617’ by inverting the /V-dimensional random permutation matrix Pi 616’ and multiplying a specific ordering vector representing the current specific ordering of the set of /V-neural network nodes of the (L-l+1)-th neural network layer with the /V-dimensional depermutation matrix Di 617’ resulting in a depermuted or descrambled ordering vector (or inverse random permutation sequence) representing a depermuted set of N neural network nodes, which is used to reconfigure the (L-l+1)-th neural network layer of the receiving DNN model. [00151] FIG. 6d illustrates an example iterative NNSI selection process 670 performed by a first device for selecting an i-th random permutation sequence for use in performing a randomization operation for randomizing the order of neural network nodes of one or more neural network layers of a transmitting DNN model, where the i-th random permutation sequence results in a reconfigured transmitting DNN model that generates, when given input communication data as input, an output communication signal that satisfies a white noise interference level when transmitted. The iterative NNSI selection process 670, performed by the first device, includes the following one or more steps of:

[00152] In step 672, selecting or generating, by the first device, an i-th random permutation sequence (or i-th randomization operation) for randomizing one or more neural network layer(s) of the transmitting DNN model. For example, the first device selects an i-th random permutation iteration from an NNSI table that maps to an indication of a white noise interference level characteristic satisfying the current white noise interference level. Alternatively, the first device generates the i-th random permutation iteration (e.g., selecting a value /) for use in generating the i-th random permutation sequence. The i-th random permutation iteration (selected or generated) is used to generate an i-th cycle of a random permutation sequence (referred to as the i-th random permutation sequence) using an initial seed and a particular random permutation function. The i-th random permutation sequence of a length N>1, which is equal to the number of neural network nodes in the set of neural network nodes of a specific neural network layer of the transmitting DNN model. The i-th random permutation sequence is used to randomly permute the set of neural network nodes of the specific neural network layer.

[00153] In step 674, reconfiguring, by the first device, the transmitting DNN model using the selected/generated the i-th random permutation sequence from step 672.

[00154] In step 676, processing, by the first device, the input communication data with the reconfigured transmitting DNN model to generate output communication signal for transmission.

[00155] In step 678, analysing and determining, by the first device, whether transmission of the generated output communication signal would satisfy the white noise interference level. This may include analysing and/or estimating a spectral density of a forecast transmission signal waveform should the first device transmit the generated output communication signal. When the generated output communication signal is determined to result in a forecast transmission signal waveform that satisfies the white noise interference level if transmitted (e.g., ‘Y’), then proceed to step 682, otherwise (e.g., ‘N’) proceed to step 680.

[00156] For example, the determination of step 678 is performed as described with reference to FIGs. 1a to 4b, in particular FIGs. 3a to 3b and 4b in relation to analysing the spectral density of the forecast transmission signal resulting from processing the output communication signal for transmission. The first device (or a component thereof) automatically analyzes the output communication signal to determine whether transmission of the output communication signal would satisfy the white noise interference level. For example, a trained spectral density estimation model processes the output communication signal for predicting or estimating the power spectral density of the transmitted signal resulting from the output communication signal. In another example, the output communication signal is input to a simulation model for generating a forecast transmission signal, where the power spectral density of the forecast transmission signal is analyzed with respect to the white interference noise level as described with reference to FIGs. 3a to 4b and/or as described herein. In another example, a simulation of a transmission of the output communication signal over a simulated communication channel is performed, where the simulation performs an analysis of whether the simulated transmission satisfies the white noise interference level and indicates the result.

[00157] In step 680, the first device updates the iteration number / to i=i+1 , and the iterative NNSI selection process 670 proceeds to step 672 for generating/selecting, by the first device, another i-th random permutation sequence. That is the generating, reconfiguring processing and analyzing steps 672, 674, 676 and 678 are repeated for the next i-th iteration.

[00158] In step 682, indicating, by the first device, the selected/generated i-th permutation sequence or i-th permutation iteration for use in reconfiguring the transmitting DNN model for generating an output communication signal that satisfies the white noise interference level when transmitted. [00159] For example, in step 682, indicating the selected/generated i-th permutation sequence includes indicating the i-th random permutation iteration for inclusion in the NNSI, where the first device sends the NNSI including the i-th random permutation iteration in a control message to the second device as described herein. Initially, the NNSI includes at least the initial seed, the random permutation function, the i-th random permutation iteration and the one or more neural network layers of the transmitting DNN model for use in generating the i-th random permutation sequence(s) for reconfiguring the transmitting DNN. Subsequent updates of NNSI may include an indication of the selected/generated i-th random permutation iteration.

[00160] In addition to selecting the i-th random permutation sequence//-//? random permutation iteration, step 672 includes selecting, by the first device, a set of one or more neural network layers, where the selected set of one or more neural network layers are different to a previous selection of one or more neural network layers. The NNSI includes the selected one or more neural network layers that result in an output communication signal that satisfies the white interference level when transmitted.

[00161] The iterative NNSI selection process 670 is performed by the first device or a controller component thereof (e.g., BS DNNC 214 of FIG. 2a), where based on the analysis performed in step 678, a mapping of the whitening characteristics of the output communication signal when transmitted and the i-th random permutation sequence//-//? random permutation iteration and/or other NNSI data (e.g., selected one or more neural network layers) may be used to populate an NNSI lookup table accessible by the first device. Further modifications include, for example, in step 672 the first device select the i-th random permutation sequence by retrieving a random permutation sequence or random permutation iteration mapped with a white noise interference characteristic that is likely to meet the white noise interference level from the NNSI lookup table. This is based on a likelihood that the corresponding whitening characteristics indicate the selected NNSI will result in a transmission of the output communication signal of the transmitting DNN satisfying the white noise interference level. In another example, the first device uses the NNSI table to bootstrap the iterative NNSI selection process 670 when searching for the e.g., i-th random permutation sequence//-//? random permutation iteration and/or other NNSI data (e.g., selected one or more neural network layers) resulting in an output communication signal that satisfies the white noise interference level when transmitted.

[00162] As an option, the one or more neural network layers of the transmitting DNN model include one or more neural network layers of the transmitting DNN model from the group of: input neural network layer, output neural network layer, one or more hidden neural network layers. The one or more neural network layers may be predefined or preselected prior to scrambling DNN operations. The selection of the one or more neural network layers may take into account the capabilities of the first and/or second devices. For example, the input neural network layer and/or output neural network layer may be selected to reduce complexity and/or computational resource consumption should the second device not have the capability to descramble hidden neural network layers. Alternatively or additionally, the one or more neural network layers may be randomly selected prior to scrambling DNN operations and/or when a new NNSI is generated for whitening the transmission signal.

[00163] FIG. 7 illustrates an example signal flow of scrambling DNN communications 720 during a communication session between a first device 104a and a second device 104b. The first and second devices 104a and 104b of FIG. 1 perform the scrambling DNN communications 720 using any of the aspects as described with reference to FIGs. 1 to 6c. For example, the first and second devices 104a and 104b perform the scrambling DNN communications 720 as the base station 210 and user equipment 220 of FIG. 2, or as the first and second devices 310 and 320 of FIGs. 3a to 3c. The signal flow of the scrambling DNN communications 720 for the communication session between the first and second device 104a and 104b include the following signal flow operations of:

[00164] In operation 721a, the first and second device 104a and 104b establish a DNN connection during a communication session between each other. During establishment of the DNN connection, the first and second devices communicate with each other for defining, agreeing, and/or configuring the type of transmitting DNN structure and receiving DNN structures for performing end-to-end communications therebetween. [00165] For example, after the first device 104a initiates a standard communication session with the second device 104b, the first device 104a selects the type of transmitting DNN structure for use in a DNN connection with the second device 104b depending on the communication channel conditions/ environment, the communication performance requirements for the DNN connection, and the type of data communications being performed (e.g. voice communication, data communications, multimedia streaming, and the like). The first device 104a requests the machine learning processing capabilities of the second device 104b to assist the selection of the transmitting DNN structure and/or receiving DNN structure for use during the DNN connection. The first device 104a accesses a DNN look-up table (or neural network table) including a set of transmitting/receiving DNN structures/models and/or pairs thereof mapped to DNN identifiers stored therein. In an example, the first device 104a stores the DNN look-up table thereon. The second device 104b accesses a corresponding DNN look-up table with a corresponding set of transmitting/receiving DNN structures/models also mapped to the same DNN identifiers. In an example, the corresponding DNN look-up table stores each receiving DNN structure/model and a mapping of the DNN identifiers corresponding to the pairs of transmitting/receiving DNN structures/models stored at the DNN lookup table accessible by the first device 104a. In an example, the second device stores the corresponding DNN look-up table thereon.

[00166] In this example, the first device 104a selects a transmitting DNN structure/model and/or corresponding receiving DNN structure/model that is suitable for the DNN connection. After selecting the transmitting DNN structure and/or corresponding receiving DNN structure, the first device 104a initiates the DNN connection by sending a DNN connection request message to the second device 104b with fields requesting establishment of a DNN connection, a DNN identifier corresponding to the selected transmitting DNN structure/receiving DNN structure and/or pair thereof, or an indication of the type of receiving DNN structure the second device 104b should use and/or the type of transmitting DNN structure the first device 104a will use so the second device 104b selects the appropriate receiving DNN structure. In some examples, when performing full duplex communications, i.e. , performing both downlink and uplink communications, the first device 104a has a transmitting DNN structure and receiving DNN structure for communications to and from the second device 104b, and the second device 104b also has a corresponding receiving DNN structure and corresponding transmitting DNN structure for communications from and to the first device 104b. The first and second devices 104a and 104b perform establishment of the DNN connection where the first device 104a configures the selected transmitting DNN structure (and/or its receiving DNN structure) for DNN communications with the second device 104b over the communication channel in one or more time slots. The second device 104b configures the corresponding receiving DNN structure (and/or its corresponding transmitting DNN structure) for DNN communications with the first device 104a over the communication channel in the one or more time slots.

[00167] In operation 721a, the white noise interference level is also initially set by the first device 104a or at the request of the second device 104b requesting improved block error rate performance (e.g., an increase in white interference noise level is requested) during establishment of the DNN connection.

[00168] In operation 721b, after a DNN connection is established with the first device 104a operating a transmitting DNN structure (and/or a receiving DNN structure) and the second device 104b operating a corresponding receiving DNN structure (and/or a corresponding transmitting DNN structure), the first device 104a performs DNN communications with the second device 104b for one or more time slots using the configured transmitting and receiving DNN structures of the first device 104a and second device 104b, respectively. In this example, the first device 104a processes input communication data using the transmitting DNN structure for transmission to the second device 104b, which processes the received transmissions using the receiving DNN structure to generate reconstructed communication data. Additionally or alternatively, the second device 104b processes input communication data using the transmitting DNN structure for transmission to the first device 104a, which processes the received transmissions using the receiving DNN structure to generate reconstructed communication data. The second device 104b sends the reconstructed communication data to a data sink of the second device 104b or sends the reconstructed communication data to one or more upper protocol layers of a protocol stack of the second device 104b. [00169] In operation 792, the first device 104a and second device 104b perform DNN scrambling operations when the first device 104a detects that the transmissions from the first device 104a does not satisfy a particular white noise interference level. For scrambling DNN communications from the first device 104a to the second device 104b (e.g., DL DNN scrambling), a TX DNN Controller of the first device 104a controls the scrambling DNN operations for the first device 104a and an RX DNN Controller of the second device 104b controls the descrambling DNN operations for the second device 104b as described with reference to FIGs. 1 to 6d, in particular FIGs. 4a to 5. For scrambling DNN communications from the second device 104a to the first device 104b (e.g., UL DNN scrambling), a TX DNN Controller of the second device 104b controls the scrambling DNN operations for the second device 104b and an RX DNN Controller of the first device 104a controls descrambling DNN operations for the first device 104a. For simplicity and by way of example only, the following steps describe scrambling DNN communications from the first device 104a to the second device 104b, where similar or the same operations are applicable for scrambling DNN communications from the second device 104a to the first device 104b, where the first device 104a swaps roles with the second device 104b.

[00170] In this example, the white noise interference level is set by the first device 104a, at the request of the second device 104b requesting improved block error rate performance (e.g, increase in white interference noise level is requested), or at the request of a third device (not shown) that is experiencing intolerable white noise interference caused by the transmissions from the first device 104a during the DNN communications. The first and second device 104a and 104b performs scrambling DNN operations in operation 792 based on, without limitation, for example the following scrambling DNN operations of:

[00171] In operation 722a, the first device 104a enables performance of the scrambling DNN operations for one or more time slots when the output communication signal from the transmitting DNN structure does not satisfy the white noise interference level when transmitted. For example, the TX DNN controller of the first device 104a detects that the transmitting DNN model of the transmitting DNN structure generates an output communication signal from input communication data that would result in a transmission signal that does not satisfy the white noise interference level. In such a case, a scrambling DNN operation is enabled and the TX DNN Controller of the first device 104a selects NNSI, as described with reference to FIGs. 1 , 2 and 3a to 6d, in particular FIGs. 3a to 6d, for use in reconfiguring the transmitting DNN model of the transmitting DNN structure (referred to as reconfiguring the transmitting DNN structure) for generating an output communication signal from the same input communication data that would result in a transmission signal satisfying the white noise interference level.

[00172] In operation 723a, the first device 104a communicates when and how scrambling DNN operations will be enabled by sending a control message to the second device 104b that includes the selected NNSI and scrambling timing information for when the receiving DNN structure at the second device 104b should be reconfigured according to the scrambling DNN operations. For example, the first device 104a sends a control message to the second device 104b with data representative of the selected NNSI and an indication of one or more time slots representing the scrambling timing information for when the first device 104a uses the selected NNSI for reconfiguring the transmitting DNN structure. Initially, the selected NNSI includes data representative of the initial seed information, type of random permutation function, specific one or more neural network layers of the transmitting DNN structure that are reconfigured for scrambling, a random permutation iteration, and any other data that enables the second device 104b to reconfigure the receiving DNN model of the receiving DNN structure (referred to as reconfiguring the receiving DNN structure) to descramble and generate reconstructed communication data. Subsequent control messages for further scrambling DNN operations include updated NNSI such as, for example, a selected random permutation iteration number, or one or more selected neural network layers that have been reconfigured for scrambling.

[00173] In operation 723b, the second device 104b acknowledges receipt (e.g., ACK) of the selected NNSI configuration and successful enablement of scrambling DNN operations at the second device 104b.

[00174] In operations 724a and 724b, the first device 104a and second device 104b turns on scrambling DNN operations based on the scrambling timing information. For example, for scrambling DNN operations from the first device 104a to the second device 104b, the first device 104a prepares to perform scrambling TX DNN operations at the first device 104a for transmitting to the second device 104b and the second device 104b prepares to perform descrambling RX DNN operations at the second device 104b for receiving transmissions from the first device 104a based on the scrambling timing information (e.g., one or more time slots). For example, the scrambling TX DNN operations includes the first device 104a reconfiguring the transmitting DNN structure based on the selected NNSI for each specific time slot associated with the scrambling timing information. The descrambling RX DNN operations includes the second device 104b reconfiguring the receiving DNN structure based on the selected NNSI for each specific time slot before the receiving DNN structure processes the transmissions arriving in each specific time slot associated with the scrambling timing information. Similarly, should there be DNN scrambling operations from the second device 104b to the first device 104a (e.g., an uplink scrambling DNN operations), the second device 104b prepares to perform similar scrambling TX DNN operations as described above with reference to the first device and the first device 104a prepares to perform descrambling RX DNN operations as described above with reference to the second device 104b.

[00175] In operation 725, the first and second devices communicate with each other using the corresponding transmitting DNN structure and receiving DNN structure and performing scrambling DNN operations based on the scrambling timing information (e.g., one or more specific time slots where scrambling is performed). For example, the first device 104a performs scrambling TX DNN operations at the first device 104a for transmitting to the second device 104b. The second device 104b performs reciprocal descrambling RX DNN operations at the second device 104b for receiving transmissions from the first device 104a based on the scrambling timing information (e.g., one or more time slots). For example, the scrambling TX DNN operations includes the first device 104a reconfiguring the transmitting DNN structure based on the selected NNSI for each specific time slot. The scrambling RX DNN operations includes the second device 104b reconfiguring the receiving DNN structure based on the selected NNSI for each specific time slot before the receiving DNN structure processes the transmissions arriving in each specific time slot. If the first device 104a and second device 104b were performing scrambling DNN communications from the second device 104b to the first device 104a (e.g., an uplink scrambling DNN operations), the second device 104b performs scrambling TX DNN operations in a similar manner as described above for the first device 104a and the first device 104b performs scrambling RX DNN operations in a similar manner as described above for the second device 104b.

[00176] Furthermore, during communications between the first device 104a and second device 104b when performing scrambling DNN operations, subsequent NNSI is iteratively determined for transmissions in one or more subsequent time slots by the first device 104a using the common/initial seed and random functions already transmitted in the first control message of operation 723a. This subsequent NNSI includes the minimum randomization information necessary for enabling the second device 104b to reconfigure the receiving DNN structure for processing transmissions received in the subsequent time slots. For example, the subsequent NNSI includes, for example, the identified random permutation sequence number or order. The first device 104a transmits the subsequent NNSI in one or more fields of a subsequent control message to the second device 104b. The subsequent NNSI also includes, for example, one or more specific neural network layers with randomization/scrambling. Subsequent scrambling DNN operations only need an updated random permutation iteration or order number when the output communication signal results in a transmission signal that does not satisfy the white noise interference level and for synchronising the reconfiguration of the receiving DNN structure of the second device 104b for processing transmissions received in the subsequent time slots. The second device 104b reuses the common/initial seed data, common seed generation, and pseudo-randomizing function indication transmitted in the initial control message.

[00177] In operation 726, the first device 104a disables scrambling DNN operations.

[00178] In operation 727a, the first device 104a sends a communication notifying the second device 104b that scrambling DNN operations are disabled/turned off.

[00179] In operation, 727b, the second device 104b acknowledges receipt (e.g., ACK) of the communication.

[00180] In operation 728a, on receipt of the acknowledgement (ACK) from the second device 104b in operation 727b, the first device 104a turns off scrambling DNN operations, and reverts the transmitting DNN structure back to its original configuration. If the first device 104a and second device 104b were performing scrambling DNN communications from the second device 104b to the first device 104a (e.g., an uplink scrambling DNN operations), where the second device 104b has a transmitting DNN structure and the first device 104b has a corresponding receiving DNN structure, then the first device 104a also reverts their receiving DNN structure back to its original configuration.

[00181] In operation 728b, on receipt of the communication for disabling scrambling DNN operations in operation 727a, the second device 104b turns off scrambling DNN operations, and reverts the receiving DNN structure back to its original configuration. If the first device 104a and second device 104b were performing scrambling DNN communications from the second device 104b to the first device 104a (e.g., an uplink scrambling DNN operations), where the second device 104b has a transmitting DNN structure and the first device 104b has a corresponding receiving DNN structure, then the second device 104b also reverts their transmitting DNN structure back to its original configuration.

[00182] In operation 729, after the first and second devices 104a and 104b disable the scrambling DNN connection, the first and second devices 104a and 104b continue to perform DNN communications with the first device 104a operating the original transmitting DNN structure (and/or a receiving DNN structure) and the second device 104b operating the original corresponding receiving DNN structure (and/or a corresponding transmitting DNN structure). For example, the first device 104a performs DNN communications with the second device 104b for one or more time slots using the configured transmitting and receiving DNNs of the first device 104a and second device 104b, respectively. In this example, the first device 104a processes input communication data using the transmitting DNN structure for transmission to the second device 104b, which processes the received transmissions using the receiving DNN structure to generate reconstructed communication data. Additionally or alternatively, the second device 104b processes input communication data using the transmitting DNN structure for transmission to the first device 104a, which processes the received transmissions using the receiving DNN structure to generate reconstructed communication data. The first and second devices 104a and 104b revert to performing a standard or conventional communications session after terminating DNN communications. [00183] FIG. 8 illustrates a signal flow of an example of scrambling DNN operations 825 between the first device 104a and second device 104b during operation 725 of the scrambling DNN communications 720 illustrated in FIG. 7. The first device 104a and second device 104b perform scrambling DNN operations 825 for communications from the first device 104a to the second device 104b of FIG. 1 using any of the aspects as described with reference to FIGs. 1 to 6c. In other example, the base station 210 and user equipment 220 of FIG. 2 perform the scrambling DNN operations 825. In another example, the first and second devices 310 and 320 of any of FIGs. 3a to 3c perform the scrambling DNN operations 825, or first and second devices 510 and 520 of FIG. 5 and the like perform the scrambling DNN operations 825. Although the first device 104a and second device 104b perform scrambling DNN operations 825, this is by way of example only and is not limiting. It is to be appreciated by the skilled person that the scrambling DNN operations 825 are applicable for communications from the second device 104b to the first device 104a, where the roles of the first and second device 104a and 104b may be reversed for some or all the operations of the scrambling DNN operations 825. In this example, the first device 104a and second device 104b establish scrambling DNN operations as described by operation 792 and, in particular, operations 722a, 723a, 723b, 724a and 724b of FIG. 7. The signal flow of the scrambling DNN operations 825 from the first device 104a to the second device 104b include the following signal flow operations of:

[00184] In operation 802, the first device 104a retrieves input communication data from a data source for input to the transmitting DNN structure of the first device 104a.

[00185] In operation 804, the transmitting DNN structure of the first device 104a processes the input communication data for generating an output communication signal for transmission in a specific time slot to the second device 104b.

[00186] In operation 806, prior to transmission of the generated output communication signal in the specific time slot, the first device 104a performs a spectral analysis to determine whether a forecast transmission of the output communication signal will satisfy a white noise interference level. For example, FIGs. 1 , 2a, 2b to 2d and/or FIGs. 4a to 4d and/or 5 and/or as describe performing spectral analysis. In response to the forecast transmission of the generated output communication signal not satisfying a white noise interference level, the operations 808 to 834 are performed, otherwise the scrambling DNN operations 825 proceeds with operation 816 for transmitting the output communication signal.

[00187] In operation 808, the first device 104a selects an updated NNSI for reconfiguring the transmitting DNN structure to generate an output communication signal that, when transmitted, satisfies the white noise interference level. The updated NNSI may be determined iteratively and/or from a NNSI lookup table as described with reference to FIGs. 1 to 6d. During selection of the updated NNSI, the transmitting DNN structure is reconfigured and reprocesses the input communication data to generate the output communication signal. After determ ining/selecti ng an updated NNSI that results in the reconfigured transmitting DNN structure generating an output communication signal that satisfies the white noise interference level, the signal flow of scrambling DNN operations 825 proceeds to operation 810.

[00188] In operation 810, the first device 104a reconfigures the transmitting DNN structure using the updated NNSI for the specific time slot.

[00189] In operation 814, the first device 104a transmits to the second device 104b the updated NNSI and the specific time slot information. For example, the first device 104a transmits a control message over a control channel to the second device 104b, the control message including an indication of the updated NNSI and specific time slot information for directing when the second device 104a is to reconfigure the receiving DNN structure of the second device 104b for processing a transmission from the first device 104a corresponding to the specific time slot information. On receipt of the control message from the first device 104a, the second device 104b performs operation 834, where the second device 104b stores the updated NNSI and specific time slot information in an NNSI buffer or NNSI table for use at the appropriate time in relation to the specific time slot information.

[00190] In operation 816, the first device 104a transmits, according to the specific time slot information, the output communication signal of the reconfigured transmitting DNN structure when transmission of the generated output communication signal satisfies the white noise interference level. [00191] On receipt of the transmission of the output communication signal in a specific time slot, the second device 104b may buffer the received output communication signal until it is ready for processing by the receiving DNN structure. In operation 848, prior to processing the received output communication signal corresponding to the specific time slot, the second device 104b reconfigures the receiving DNN structure using the updated NNSI for the specific time slot. The second device 104b retrieves the updated NNSI for the specific time slot from storage (e.g., from the NNSI buffer or NNSI table) and uses it to reconfigure the receiving DNN structure as described, for example, with reference to FIGs. 1 to 6d, in particular FIGs. 3a to 3c, 4c and 4d, and 5 and/or as described herein. The reconfiguration of the receiving DNN structure modifies the receiving DNN structure for performing the reverse operations performed by the reconfigured transmitting DNN structure, i.e., to descrambling operations, to generate reconstructed communication data corresponding to the input communication data transmitted in the specific time slot.

[00192] In operation 850, the receiving DNN structure processes the received output communication signal for the specific time slot and generates reconstructed communication data corresponding to the input communication data transmitted in the specific time slot. The second device 104b sends the reconstructed communication data to a data sink at the second device 104b or sends the reconstructed communication data to one or more upper protocol layers of a protocol stack of the second device 104b.

[00193] In operation 852, on successful generation of the reconstructed communication data, the second device 104b sends an acknowledgement to the first device 104a indicating successful receipt of the input communication data for the specific time slot. In operation 853, the second device 104b proceeds to prepare for receiving a next transmission from the first device 104a in one or more subsequent time slots.

[00194] In operation 825a, the first device 104a on receipt of the acknowledgement in operation 852, proceeds to the next transmission of, if any, further input communication data from the data source in one or more subsequent time slots, where the scrambling DNN operations 825 repeats until it is determined to disable scrambling DNN communications as described in operation 726 of FIG. 7. [00195] FIG. 9 illustrates a signal flow diagram of another example of scrambling DNN operations 925 between the first device 104a and second device 104b during operation 725 of the scrambling DNN communications 720 illustrated in FIG. 7. The scrambling DNN operations 925 further modifies the operations of scrambling DNN operations 825 of FIG. 8 by the insertion of operation 917, where the first device 104a identifies, in between transmissions, further NNSI that satisfies the white noise interference level. The signal flow of the scrambling DNN operations 925 from the first device 104a to the second device 104b include the following signal flow operations of:

[00196] At the first device, operations 902-916 substantially correspond to operations 802-816 in FIG. 8. Similarly, at the second device, operations 934, 948, 950, 952 and 953 substantially correspond to operations 834, 848, 850, 852 and 853 in FIG. 8. At the first device 104a, after operation 916, the first device 104a performs operation 917, where the first device 104a continues to identify further NNSI to assist with reconfiguring the transmitting DNN structure for generating an output communication signal that is forecast to satisfy the white noise interference level when transmitted. The first device 104a stores any identified NNSI mapped to an NNSI identifier and the associated or estimated white noise interference level in an NNSI look-up table for use in operation 908 (or operation 808 of FIG. 8). This may bootstrap the search for updated NNSI in operation 908 providing faster selection of updated NNSI that would result in the generated output communication signal satisfying the white noise interference level when transmitted. In an example, the first device 104a performs operation 917 in the background and/or in between operations 916 and 925a.

[00197] In operation 925a, the first device 104a on receipt of the acknowledgement in operation 952, proceeds to the next transmission of, if any, further input communication data from the data source in one or more subsequent time slots, where the scrambling DNN operations 925 repeats until it is determined to disable scrambling DNN communications as described in operation 726 of FIG. 7.

[00198] As an option, the first device 104a sends the second device 104b updates of any identified NNSI, NNSI identifier and/or associated or estimated white noise interference level thereto via one or more control messages for use by the second device 104b in updating its NNSI look-up table with the NNSI and NNSI identifier and the like. This provides the advantage that for a selected NNSI by the first device 104a, the control message for reconfiguring the receiving DNN structure of the second device 104b can send the NNSI identifier of the selected NNSI and timing information, where the second device 104b can retrieve the selected NNSI accordingly.

[00199] FIG. 10 illustrates a signal flow of another example of scrambling DNN communications 1000 between a first device 104a, a second device 104b and a third device 1003. The scrambling DNN communications 1000 further modifies the scrambling DNN communications 720 of FIG. 7 by insertion of operations 1005a and 1005b, where the third device 1003 communicates with the first device 104a in relation to a tolerable white noise interference level. The signal flow of the scrambling DNN communications 1000 from the first device 104a to the second device 104b includes the following signal flow operations of:

[00200] At the first and second devices 104a and 104b, operations 1021a, 1021 b, 1092, 1026, 1027a, 1027b, 1028a, 1028b, and 1029 substantially correspond to operations 721a, 721b, 792, 726, 727a, 727b, 728a, 728b, and 729 as described with reference to FIG. 7. In operation 1021a, the white noise interference level is initially set by the first device 104a or at the request of the second device 104b requesting improved block error rate performance (e.g., an increase in white interference noise level is requested) during establishment of the DNN connection. After the first and second devices 104a and 104b have established a connection for DNN communications and are performing DNN communications in operation 1021 b therebetween, the third device 1003 may experience intolerable white noise interference caused by the transmissions from the first device 104a (or second device 104b) during the DNN communications in operation 1021 b.

[00201] In operation 1005a, the third device 1003 sends a notification to the first device 104a (or second device 104b) indicating that transmissions of output communication signals from the first device 104a to the second device 104b exceed an acceptable white noise interference level associated with the third device 1003. In this example, the third device 1003 is a victim device that is experiencing interference from transmissions of the first device 104a. In an example, when the first device 104a is a base station, the third device 1003 contacts the first device 104a via an uplink channel. In another example, the third device 1003 is also a base station of another cell, and requests via the core network that the first device 104a reduces the interference caused to other user equipment within its cell. In an example, the notification from the third device 1003 indicates a tolerable white noise interference level or an acceptable white noise interference level. In another example, the notification from the third device 1003 indicates that the current white noise interference level output by the first device 104a is not acceptable.

[00202] In operation 1005b, the first device 104a, on receiving the notification from the third device 1003 in operation 1005a, the first device 104a adjusts the white noise interference noise level by decreasing the white noise interference level. In an example, if the notification from the third device 1003 indicates a tolerable white noise interference level or an acceptable white noise interference level, then the first device 104a adjusts the white interference noise level to the tolerable or acceptable white noise interference level. In another example, if the notification from the third device 1003 indicates that the current white noise interference level output by the first device 104a is not acceptable, then the first device 104a adjusts the white noise interference level by incrementally decreasing the white noise interference level or until the third device 1003 ceases the notifications for reducing the white noise interference level. For example, the third device 1003 (e.g., a victim UE) moves away from the first device 104a such that it does not experience an intolerable white noise interference level. In another example, as the third device 1003 moves away from the first device 104a, it sends further notifications to the first device 104a indicating further tolerable white noise interference levels. These further white interference noise levels may increase over previous white noise interference levels as the distance the third device 1003 moves away from the first device 104a increases.

[00203] The first and second devices 104a and 104b perform operation 1092 with the adjusted white noise interference level. The remaining operations 1026, 1027a, 1027b, 1028a, 1028b, and 1029 of FIG. 10 substantially correspond to the operations 726, 727a, 727b, 728a, 728b, and 729 of FIG. 7, respectively.

[00204] FIG. 11 illustrates a signal flow diagram of another example of scrambling DNN operations 1125 between the first device 104a and second device 104b during operation 725 or 1092 of the scrambling DNN communications 720 or 1000 illustrated in FIGs. 7 or 10. The scrambling DNN operations 1125 further modifies the operations of scrambling DNN operations 825 of FIG. 8 or 925 of FIG. 9 by the insertion of operations 1105a and 1105b, where the first device 104a is notified by a third device 1103 to adjust the white noise interference level to an acceptable level. The signal flow of the scrambling DNN operations 1125 from the first device 104a to the second device 104b include the following signal flow operations of:

[00205] At the first device 104a, the operations 1102, 1104, 1106, 1108, 1110, 1114, 1116, and 1125a substantially correspond to the operations 802, 804, 808, 810, 814, 816, and 825a of FIG. 8, or the operations 902, 904, 908, 910, 914, 916, and 925a of FIG. 9. At the second device 104b, the operations 1134, 1148, 1150, 1152, and 1153 substantially corresponding with operations 834, 848, 850, 852, and 853 described with reference to FIG. 8, or operations 934, 948, 950, 952, and 953 described with reference to FIG. 9. After the first and second devices 104a and 104b perform scrambling DNN operations 1125, first and second devices 104a and 104b perform operations 1102, 1104, 1106, 1108, 1110, 1114, and 1116 and operations 1134, 1148, 1150, 1152, and 1153, respectively, for one or more time slots. During these one or more time slots, a third device 1103 experiences an intolerable white noise interference level despite the scrambling DNN communications between first and second devices 104a and 104b in which the transmitting DNN structure of the first device 104a generates an output communication signal that satisfies a current white noise interference level when transmitted. This means the current white noise interference level set by the first device 104a is too high and the transmissions therefrom still result in intolerable or an unacceptable level of interference at the third device 1103.

[00206] As described in operation 1005a of FIG. 10, in operation 1105a, the third device 1103 sends a notification to the first device 104a (or second device 104b) indicating that transmissions of output communication signals from the first device 104a to the second device 104b are exceeding an acceptable white noise interference level associated with the third device 1103. As described in operation 1005b of FIG. 10, in operation 1105b, the first device 104a, on receiving the notification from the third device 1103 in operation 1105a, adjusts the white noise interference noise level by decreasing the white noise interference level to the acceptable amount if indicated by the third device 1103, or an incremental amount. If the latter, the third device 1103 sends further notifications indicating the adjusted white noise interference level is still intolerable or unacceptable, where the first device 104a adjusts by a further incremental amount until the third device 1103 ceases to send any further notifications or sends a notification indicating the white noise interference level is acceptable. The further examples as described for operations 1005a and 1005b are also applicable to operations 1105a and 1105b of FIG. 11.

[00207] In operation 1125a, the first device 104a on receipt of the acknowledgement in operation 1152 from second device 104b, proceeds to the next transmission of, if any, further input communication data from the data source in one or more subsequent time slots, where the signal flow of the scrambling DNN operations 1125 repeats until it is determined to disable scrambling DNN communications as described in operations 726 of FIG. 7 or 1026 of FIG. 10.

[00208] FIG. 12 illustrates a signal flow of an example of scrambling DL/UL DNN communications 1220 during a DL/UL DNN communication session between a BS 210 and a UE 220. In this example, reference numerals of FIG. 2a are used for the same or similar components. The scrambling DNN communications 720 and 1020 are further modified to include scrambling DL/UL DNN communications 1220 between the BS 210 and UE 220 during a DL/UL DNN communication session therebetween. The BS 210 and UE 220 of FIG. 2 perform scrambling DL/UL DNN communications 1220 using any of the aspects as described with reference to FIGs. 1 to 11 . In particular, at the BS 210 and UE 220, the operations 1221a, 1221 b, 1292, 1226, 1227a, 1227b, 1228a, 1228b, and 1229 further modify the corresponding operations 721a, 721b, 792, 726, 727a, 727b, 728a, 728b, and 729 as described with reference to FIG. 7, and/or the corresponding operations 1021a, 1021 b, 1092, 1026, 1027a, 1027b, 1028a, 1028b, and 1029 of FIG. 10 for use with a DL/UL DNN communication session between BS 210 and UE 220. The signal flow of the scrambling DL/UL DNN communications 1220 for the communication session between the BS 210 and UE 220 include the following signal flow operations of:

[00209] In operation 1221a, the BS and UE perform a radio resource control (RRC) DNN connection establishment for establishing a DL/UL DNN communication session between each other. The DL/LIL DNN communication session includes DL DNN communications from BS 210 to UE 220, and UL DNN communications from UE 220 to BS 210. During establishment of the DL/UL DNN communication session, the BS 210 and UE 220 communicate with each other for defining, agreeing, and/or configuring the type of DL/UL transmitting DNN structure and DL/UL receiving DNN structures that each will use in DL DNN communications and UL DNN communications for performing end-to-end communications therebetween.

[00210] For example, when UE 220 is in an RRCJDLE state, then the UE 220 sends an RRC DNN connection request for establishing a DL/UL DNN communication session to the BS 210. The RRC DNN connection request may include the UE identity and capabilities of the UE to enable the BS 210 to select the appropriate DL/UL transmitting DNN structure and DL/UL receiving DNN structures for DL and UL communication channels (e.g., Physical Downlink Shared Channel (PDSCH) and Physical Uplink Shared Channel (RUSCH)), whilst also defining the control channels (e.g., Physical Downlink Control Channel (PDCCH) or Physical Uplink Control Channel (PUCCH)) for use in DL and/or UL DNN communications between UE 220 and BS 210. The BS 210 has set of DL transmitting/receiving DNN structures and/or pairs thereof and a set of UL transmitting/receiving DNN structures and/or pairs thereof each of which are mapped to DL I UL DNN identifiers, respectively, and stored in a DNN look-up table at the BS 210. The UE 220 has a corresponding set of UL I DL transmitting/receiving DNN structures also mapped to the same UL I DL DNN identifiers and stored in a DNN look-up table at the UE 220. The BS 210 selects DL transmitting and receiving DNN structures and/or UL transmitting and receiving DNN structures based on the DL and/or UL communication channel I environment, the communication performance requirements for the DL and/or UL DNN connections, and the type of DL and/or UL data communications (e.g., voice communication, data communications, multimedia streaming, and the like). The BS 210 sends a RRC DNN connection setup message including data representative of the selected DL/UL receiving DNN structures (e.g., DL/UL DNN identifiers) for DL and UL communication channels (e.g., PDSCH and PUSCH), the DL and UL control channels (e.g., PDCCH/PUCCH) for use in DL and UL DNN communications between UE 220 and BS 210. The UE 220 on receipt of the RRC DNN connection setup message may use the DL DNN identifier and UL DNN identifier on the DNN look-up table to select and configure the corresponding DL receiving DNN structure and UL transmitting DNN structure. The UE 220 may send a RRC connection setup complete message to the BS 210 indicating the UE 220 has configured the corresponding DL receiving DNN structure and UL transmitting DNN structures accordingly and is ready for DL and UL DNN communications with BS 210. The BS 210 also configures the corresponding DL transmitting DNN structure and UL receiving DNN structure for DL and UL DNN communications with UE 220.

[00211] Alternatively, when the UE 220 is in an RRC_CONNECTED state, the BS 220 sends an RRC DNN connection reconfiguration message to the UE 220 for modifying an existing RRC connection into an RRC DNN connection including data representative the appropriate DL/UL transmitting DNN structures and DL/UL receiving DNN structures for the DL and UL communication channels (e.g., PDSCH and PUSCH), whilst also defining the control channels (e.g., PDCCH/PUCCH) for use in DL and UL communications between UE 220 and BS 210. The UE 220 may send a RRC reconfiguration complete message to the BS 210 indicating the UE 220 has configured the corresponding DL receiving DNN structure and UL transmitting DNN structures accordingly and is ready for DL and UL DNN communications with BS 210. The BS 210 also configures the corresponding DL transmitting DNN structure and UL receiving DNN structure for DL and UL DNN communications with UE 220.

[00212] In operation 1221a, during establishment of the DL/UL DNN connection or afterwards, the BS 210 sets the white noise interference level for the DL communication channel and/or UL communication channel to a default setting, or to a certain white noise interference level depending on the performance requirements of the DL/UL DNN communication session. Alternatively or additionally, in an example, the UE 220 requests a certain white noise interference level to improve block error rate performance for DL and/or UL DNN communications (e.g., an increase in white interference noise level is requested).

[00213] In operation 1221b, after the BS 210 and UE 220 establish a DL/UL DNN communication session, the BS 210 and UE 220 perform DL DNN communications for one or more time slots using the configured DL transmitting and DL receiving DNN structures, respectively. The UE 220 and BS 210 also perform UL DNN communications for one or more time slots using the configured UL transmitting and UL receiving DNN structures, respectively.

[00214] In operation 1292, The BS 210 and UE 220 performs DL and/or UL DNN scrambling when the BS 210 detects that the transmissions from the BS 210 does not satisfy the particular white noise interference level for DL DNN communications, or the transmissions from the UE 220 does not satisfy a particular white noise interference level for UL DNN communications. For DL DNN scrambling communications from the BS 210 to the UE 220, a TX DNN Controller of the BS 210 and a RX DNN Controller of the UE 220 control the scrambling DL DNN operations as described with reference to FIGs. 1 to 11 , in particular FIGs. 4a to 5. Similarly, for UL DNN scrambling communications from the UE 220 to BS 210, a TX DNN Controller of the UE 220 and an RX DNN Controller of the BS 210 control the scrambling UL DNN operations.

[00215] As described, the white noise interference level is set by the BS 210, at the request of the UE 220 regarding improved block error rate performance, or at the request of another UE or BS (e.g., see FIGs. 10 or 11 in relation to third device 1003 or 1103) experiencing intolerable white noise interference caused by the transmissions from the BS 210 during the DL DNN communications or the UE 220 during UL DNN communications. The BS 210 and UE 220 perform DNN operations 1292 on DL/UL based on, without limitation, for example the following scrambling DL/UL DNN operations of:

[00216] In operation 1222a, the BS 210 enables performance of the scrambling DL DNN operations for one or more time slots when the output communication signal from the DL transmitting DNN structure of the BS 210 is forecast to not satisfy the white noise interference level for the DL communication channel if transmitted over the DL communication channel (e.g., PDSCH). For example, the DL TX DNN controller of the BS 210 detects that the DL transmitting DNN structure generates an output communication signal from input communication data forecast to result in a transmission signal on the PDSCH that does not satisfy the white noise interference level. In such a case, a DL scrambling DNN operation is enabled and the DL TX DNN controller of the BS 210 selects NNSI, as described with reference to FIGs. 1 , 2 and 3a to 6d, in particular FIGs. 3a to 6d, for use in reconfiguring the DL transmitting DNN structure for generating an output communication signal from the same input communication data that would result in a transmission signal over the PDSCH satisfying the white noise interference level. Similarly, UL scrambling DNN operations are enabled when the UE 220 or the BS 210 detects that the transmission signals over the UL communication channel (e.g., PUSCH) do not satisfy the white noise interference level set for the UL communication channel. The BS 210 selects NNSI for use in reconfiguring the UL transmitting DNN structure for use by the UE 220 in generating an output communication signal therefrom that would result in a transmission signal over the PUSCH satisfying the white noise interference level.

[00217] In operation 1223a, the BS 210 communicates when and how scrambling DL and UL DNN operations will be enabled by sending, for example, an RRC control message (e.g., RRC Connection Reconfiguration), an Medium Access Control (MAC) message, or Downlink Control Information (DCI) message to the UE 220 over the PDCCH with fields including the selected NNSI configuration and also DL and/or UL timing information for when the DL receiving DNN structure at the UE 220 should be reconfigured and/or UL transmitting DNN structure at the UE 220 should be reconfigured. Initially, the selected NNSI includes data representative of the initial seed information, type of random permutation function, specific layers of the DL and/or UL transmitting DNN structure that are to be reconfigured, random permutation iteration, and any other data that enables the UE 220 to reconfigure the DL receiving DNN structure for generating reconstructed communication data and/or reconfiguring the UL transmitting DNN structure for generating output communication signal for transmission over the PUSCH. Subsequent RRC/MAC/DCI messages for further DL or UL scrambling DNN operations include updated NNSI configurations such as, for example, a selected random permutation iteration number, or one or more selected neural network layers that have been reconfigured for DL and/or UL scrambling.

[00218] In operation 1223b, the UE 220 acknowledges receipt (e.g., ACK) of the selected NNSI configuration for scrambling/descrambling DL DNN operations (e.g., descrambling DL RX DNN operations performed by UE 220) and/or scrambling UL DNN operations (e.g., scrambling UL TX DNN operations performed by UE 220), and enablement of scrambling DL/UL DNN operations. [00219] In operations 1224a and 1224b, the BS 210 and UE 220 turns on scrambling DL and/or UL DNN operations based on the DL and/or UL timing information. For example, the BS prepares to perform scrambling DL TX DNN operations for DL transmission to the UE 220 whilst the UE 220 prepares to perform descrambling DL RX DNN operations for receiving the DL transmissions from the BS 210 using the DL timing information (e.g., one or more DL time slots). In another example, the UE 220 prepares to perform scrambling UL TX DNN operations for transmitting UL transmissions from the UE 220 to the BS 210 whilst the BS 210 prepares to perform descrambling UL RX DNN operations for UL transmissions from the UE 220 using the UL timing information (e.g., one or more UL time slots).

[00220] In operation 1225, the BS 210 and UE 220 perform scrambling DL/UL DNN operations in communications with each other over PDSCH and/or PUSCH in a similar manner as described with reference to FIGs. 1 to 11 , in particular FIGs. 7 and 10.

[00221] In operation 1226, after scrambling DL/UL DNN operations/communications are determined to be unnecessary, the BS 210 disables scrambling DL/UL DNN operations.

[00222] In operation 1227a, when the BS 210 disables or turns off scrambling DL and/or UL DNN operations, the BS 210 sends to the UE 220, for example, an RRC message, Medium Access Control (MAC) message, or DCI message notifying the UE 220 that scrambling DL and/or UL DNN operations are being disabled/turned off.

[00223] In operation, 1227b, the UE 220 acknowledges receipt (e.g., ACK) of the RRC/MAC/DCI message that the BS 210 sent in operation 1227a.

[00224] In operation 1228a, on receipt of the acknowledgement (ACK) from the UE 220 in operation 1227b, the BS 210 turns off scrambling DL / UL DNN operations, and reverts the DL transmitting DNN structure and/or the UL receiving DNN structure back to its original configuration. Alternatively, the BS 210 retains the current DL transmitting DNN structure and/or the UL receiving DNN structure (e.g., the latest reconfiguration), which still allows any further DNN communications to likely satisfy the white noise interference level. [00225] In operation 1228b, on receipt of the RRC/DCI message for disabling scrambling DL/UL DNN operations in operation 1227a, the UE 220 turns off scrambling DL and/or UL DNN operations, and reverts the receiving DL DNN structure and/or the UL transmitting DNN structure back to its original configuration. Alternatively, if the BS 210 has also indicated that it will do so, the UE 220 retains the current DL receiving DNN structure and/or the current UL transmitting DNN structure (e.g., the latest reconfiguration), which still allows any further DNN communications to likely satisfy the white noise interference level.

[00226] In operation 1229, after the BS 210 and UE 220 disable the scrambling DNN connection, the BS 210 and UE 220 continue to perform DNN communications. The BS 210 and UE 220 revert to performing, for example, a standard or conventional 3G to 4G, 5G or 6G communications session therebetween when DNN communications are not needed.

[00227] FIG. 13 illustrates a signal flow of an example of scrambling DL DNN operations 1325 between the BS 210 and UE 220 during operation 1225 in the scrambling DL/UL DNN communications 1220 of FIG. 12. In this example, reference numerals of FIG. 2a are used for the same or similar components. The example scrambling DNN operations 825 are further modified to include scrambling DL DNN operations between the BS 210 and UE 220 during a DL DNN communication session therebetween. The BS 210 and UE 220 of FIG. 2a perform scrambling DL DNN operations 1325 using any of the aspects as described with reference to FIGs. 1 to 12. In particular, at the BS 210, operations 1302-1316 substantially correspond to the operations 802-816 as described with reference to FIG. 8 but modified for scrambling DL DNN communications. Similarly, at the UE, operations 1334, 1348, 1350, 1352 and 1353 substantially correspond to operations 834, 848, 850, 852 and 853 as described with reference to FIG. 8 but modified for scrambling DL DNN communications. In this example, DNN operation 1292 of FIG. 12 configures the BS 210 and UE 220 to perform scrambling DL DNN operations 1325. The signal flow of the scrambling DL DNN operations 1325 from the BS 210 to UE 220 include the following signal flow operations of:

[00228] In operations 1302 to 1306 substantially correspond to operations 802 to 804 of FIG. 8 except that the DL transmitting DNN structure of the BS 210 processes the input communication data for generating an output communication signal for transmission over the PDSCH in a specific time slot to the UE 220.

[00229] Operation 1306 further modifies operation 806 of FIG. 8 by the BS 210 communicating any updated NNSI and the specific time slot information to the UE 220, for example, using RRC/DCI messaging over PDCCH. On receipt of the RRC/DCI messaging from the BS 210, the UE 220 performs operation 1334 in a similar manner as operation 834 of FIG. 8, where an NNSI buffer or NNSI table accessible by UE 220 stores the updated NNSI and specific time slot information for use in descrambling received transmissions over PDSCH in relation to the specific time slot information.

[00230] In operation 1316, the BS 210 transmits, according to the specific time slot information, the output communication signal of the reconfigured DL transmitting DNN structure on the PDSCH when transmission of the generated output communication signal satisfies the white noise interference level. On receipt of the transmission over PDSCH of the output communication signal in a specific time slot, the UE 220 may buffer the received output communication signal until it is ready for processing by the DL receiving DNN structure.

[00231] Operations 1348 and 1350 substantially correspond to operations 848 and 850 except that the DL receiving DNN structure is reconfigured using the updated NNSI for the specific time slot, and the DL receiving DNN structure processes the received output communication signal for the specific time slot and generates reconstructed communication data corresponding to the input communication data transmitted in the specific time slot. The UE 220 sends the reconstructed communication data to a data sink at the UE 220 or sends the reconstructed communication data to one or more upper protocol layers of a protocol stack of the UE 220.

[00232] Operations 1352 and 1325a correspond substantially to operations 852 and 825a of FIG. 8. The scrambling DL DNN operations 1325 repeats until it is determined to disable scrambling DL/UL DNN communications as described in operation 1226 of FIG. 12. [00233] FIG. 14 illustrates a signal flow of an example of scrambling UL DNN communications 1420 between the UE 220 and BS 210. In this example, reference numerals of FIG. 2a are used for the same or similar components. In this example, operations 1421a/b, 1422a, 1423a, 1423b, 1424a, and 1424b correspond to operations 1221a/b, 1222a, 1223a, 1223b, 1224a, and 1224b except that the UE 220 performs scrambling UL DNN communications over PUSCH using an UL transmitting DNN structure (e.g., UL TX DNN) and the BS 210 performs scrambling UL DNN communications over PUSCH using a UL receiving DNN structure (e.g., UL RX DNN).

[00234] When the UE 220 has input communication data for transmission, the UE 220 uses the UL transmitting DNN structure to generate an UL output communication signal for transmission on PUSCH. Operation 1406 substantially corresponds to operations 806 or 1306 of FIGs. 8 or 13 except that the UE 220 performs these operations, where the UE 220 instead detects whether a forecast transmission of the UL output communication signal does not satisfy the white noise interference level for the PUSCH. On detecting the white interference noise level is not satisfied, the UE 220 performs operations 1408, 1410, 1414 based on the following:

[00235] In operation 1408 substantially corresponds to operations 808 or 1308 except the UE 220 selects an updated NNSI for reconfiguring the UL transmitting DNN structure to generate an UL output communication signal that, when transmitted on PUSCH, satisfies the white noise interference level in the specific time slot. This may be iteratively selected as described as described with reference to FIGs. 1 to 6d and/or operations 808 or 1308 of FIGs. 8 or 13. After the UE 220 selects the updated NNSI, the UE 220 proceeds to operation 1410.

[00236] Operations 1410 and 1414 substantially corresponds to operations 810 and 814 or 1310 or 1314 except the UE 220 reconfigures the UL transmitting DNN structure using the updated NNSI for the specific time slot and the UE 220 communicates using uplink control signalling on PUCCH the updated NNSI to the BS 210. The BS 210 already has the specific time slot information, which is determined by the BS 210 in RRC/DCI messaging to the UE 220. On receipt of the uplink control signalling, the BS 210 performs operation 1434, which substantially corresponds to operations 834 or 1334 except the BS 210 stores the updated NNSI with the corresponding specific time slot in an NNSI buffer or NNSI table at the BS 210 for descrambling received transmissions on PUSCH using the UL receiving DNN structure in relation to the specific time slot information.

[00237] In operation 1416, the UE 220 transmits, according to the specific time slot, the UL output communication signal of the reconfigured UL transmitting DNN structure on the PUSCH when transmission of the generated UL output communication signal satisfies the white noise interference level.

[00238] Operations 1448 and 1450 substantially correspond to operations 848 and 850 of FIG. 8 or 1348 and 1350 of FIG. 13 except that the BS 210 performs these operations on receipt of the transmission on PUSCH of the UL output communication signal in the specific time slot, which the BS 210 buffers prior to processing the received UL output communication signal for the specific time slot using the UL receiving DNN structure reconfigured using the updated NNSI for the specific time slot. The UL receiving DNN structure processes the received UL output communication signal for the specific time slot and generates reconstructed communication data corresponding to the input communication data transmitted by UE 220 on PUSCH in the specific time slot. The BS 210 sends the reconstructed communication data to a data sink at the BS 210 or sends the reconstructed communication data to one or more upper protocol layers of a protocol stack of the BS 210.

[00239] Operations 1452 and 1453 substantially corresponds to operations 852 and 853 of FIG. 8 or operations 1352 or 1353 of FIG. 13 except it is performed by the BS 210.

[00240] In operation 1420a, the UE 220 on receipt of the acknowledgement in operation 1452, proceeds to the next transmission on the PUSCH of, if any, further input communication data from the data source at UE 220 in one or more subsequent time slots, where the scrambling UL DNN communications 1420 repeats until it is determined to disable scrambling UL DNN communications as described in operation 1226 of FIG. 12.

[00241] FIG. 15 illustrates a signal flow of another example of scrambling UL DNN communications 1520 between the UE 220 and BS 210. In this example, reference numerals of FIG. 2a are used for the same or similar components. In this example, operations 1521a/b, 1522a, 1523a, 1523b, 1524a, and 1524b correspond to operations 1421 a/b, 1422a, 1423a, 1423b, 1424a, and 1424b. When the UE 220 has input communication data for transmission it uses the UL transmitting DNN structure to generate an UL output communication signal for transmission on PUSCH. Operation 1506 further modifies operations 1406 of FIG. 14 based on the following modifications:

[00242] In operation 1507, after detecting the UL output communication signal does not satisfy the white interference noise level if transmitted, the UE 220 communicates using uplink control signalling on PUCCH a request for updated NNSI from the BS 210.

[00243] Operation 1508 substantially corresponds to operations 808 or 1308 of FIGs. 8 or 13, respectively, except the BS 210 selects an updated NNSI for reconfiguring the UL transmitting DNN structure of the UE 220 for generating an UL output communication signal that, when transmitted on PUSCH, satisfies the white noise interference level in the specific time slot. This may be iteratively selected by the BS 210 as described with reference to FIGs. 1 to 6d and/or operations 808, 1308 and 1408 of FIGs. 8, 13, and 14. In an example, as described with reference to FIG. 2a, when the BS 210 selects the NNSI for the UE 220, the BS 210 simulates the UL with the random UE UL input communication data and UE UL transmitting DNN structure to generate an output communication signal, and selects the NNSI that results in the output communication signal satisfying the white noise interference level for the UL.

[00244] Operation 1514 substantially corresponds to operation 814 and 1314 except the BS 210 communicates the updated NNSI for the UL and the specific time slot information to the UE 220, for example, using RRC/DCI messaging over PDCCH. In operation 1534a, the BS 210 stores the updated NNSI and specific time slot information in an NNSI buffer or NNSI table at the BS 210 for use in descrambling the received transmission on PUSCH in relation to the specific time slot information. As well, on receipt of the RRC/DCI messaging from the BS 210, the UE 220 performs operation 1534b, and stores the updated NNSI and specific time slot information in an NNSI buffer or NNSI table at the UE 220 for use in scrambling the input communication data for transmission on PLISCH according to the specific time slot information. In operation 1510, the UE 220 reconfigures the UL transmitting DNN structure of the UE 220 using the updated NNSI for the specific time slot.

[00245] In operation 1516 substantially corresponds to operations 816 and 1316 of FIGs. 8 or 13, respectively, except the UE 220 transmits, according to the specific time slot, the UL output communication signal on the PUSCH accordingly.

[00246] Operations 1516, 1548, 1550, 1552, 1554 and 1520a correspond to operations 1416, 1448, 1450, 1452, 1454 and 1420a of FIG. 14. The UE 220 proceeds to the next transmission on the PUSCH of, if any, further input communication data from the data source at UE 220 in one or more subsequent time slots, where the scrambling UL DNN communications 1520 repeats until it is determined to disable scrambling UL DNN communications as described in operation 1226 of FIG. 12.

[00247] FIG. 16 illustrates signal flow of a scrambling DL/UL DNN communication session 1600 between a BS 210 and a UE 220 as described with reference to FIG. 2a and FIGs. 12 to 15. In this example, reference numerals of FIG. 2a are used for the same or similar components. The BS 210 includes a BS DNN Controller 214 (BS DNNC), a BS DL transmitting DNN structure (BS DL TX DNN) for processing and scrambling input communication data for downlink transmission over PDSCH to UE 220, and a BS UL receiving DNN structure (BS UL RX DNN) for processing received uplink transmissions on PUSCH from UE 220. The UE 220 includes a UE DNN Controller 224 (UE DNNC), a UE DL receiving DNN structure (UE DL RX DNN) for processing received downlink transmissions over PDSCH from BS 210, and a UE UL transmitting DNN structure (UE UL TX DNN) for uplink transmissions on PUSCH to BS 210. It is also assumed that the BS 210 has assigned the UE 220 appropriate frequency/time slots for control plane signalling over corresponding PDCCH and PUCCH.

[00248] In operation 1621a, the BS DNNC 214 establishes a UL/DL DNN communications session between BS 210 and UE 220. The BS DNNC 214 selects the DL transmitting and receiving DNN structure (BS DL TX DNN and UE DL RX DNN) pair for use in the scrambling DL/UL DNN communication session 1600 in which the BS 210 transmits an RRC Establishment Request message including a DL DNN type or identifier associated with the selected DL TX/RX DNN pair (e.g., RRC DNN Establishment Request (DL DNN Type/ld)). The BS DNNC 214 retrieves the TX DNN configuration data for the BS DL TX DNN 228 corresponding to selected DL DNN type or identifier from BS TX/RX DNN store or table 215b. In operation 1621b, the BS DNNC 214 sends a configuration instruction (e.g., Cfg(DLDNN Type)) to configure the BS DL TX DNN structure 206. In operation 1621c, when the UE DNNC 224 receives the RRC Establishment Request message including the DL DNN type or identifier associated with the selected DL TX/RX DNN pair, the UE DNNC 224 retrieves the RX DNN configuration data for the UE DL RX DNN 228 corresponding to the received DL DNN type or identifier from UE TX / RX DNN storage / table 225b at the UE 220, and sends a configuration instruction (e.g., Cfg(DL DNN Type)) to UE DL RX DNN structure 228 to configure the UE DL RX DNN 228 based on the retrieved RX DNN configuration data. In operation 1621d, after configuration of the UE DL RX DNN 228, the UE DNNC 224 of the UE 220 sends an RRC response indicating acknowledgement to configuring the UE DL RX DNN 228 to the BS 210 (e.g., RRC DNN Establishment Resp (ACK)).

[00249] On receipt of the acknowledgement, in operation 1602i, input communication data for transmission in a specific time slot, TS / (e.g. , l_Data Xi) is applied to the BS DL TX DNN 206, which generates a DL output communication (OC) signal corresponding to l_Data Xi (e.g., OC_Data Xi). In operation 1604i, the DL OC_Data Xi is provided to the BS DNNC 214 for transmission as a transmit waveform signal over PDSCH to the UE 220 in TS /. As described with reference to FIG. 5, on receiving OC_Data Xi, the BS DNNC 214 detects whether transmission of OC_Data Xi would satisfy a white interference noise level, if this is the case, then the OC_Data Xi is, for example, buffered and transmitted to the UE 220 as a transmission waveform signal over PDSCH in the specific TS /. Otherwise, NNSI is generated/selected to reconfigure DL TX DNN 206 to generate an OC_Data Xi that satisfies the white interference noise level as described with reference to FIG. 5 and/or any of FIGs. 1 to 4d and/or 6a to 15. In operation 1616i, the BS DNNC 214 transmits the OC_Data Xi to the UE 220 as a transmission waveform signal in PDSCH in TS / (e.g., PDSCH RF TX WAVEFORM (OC_DataXi, TS /)). The UE 220 receives the transmission signal waveform over PDSCH in TS / and processes (e.g., down converts) into received OC_Data Xi (e.g., Rx OC Xi)), and in operation 1650i-1 , the Rx OC Xi for TS / is input to the UE DL RX DNN 228 for processing. In operation 1650i-2, the UE DL RX DNN 228 processes the Rx OC Xi and generates reconstructed communication data for TS / corresponding to the l_Data Xi (e.g., RJData Xi). In operation 1652i , after the UE DL RX DNN 228 has successfully generated RJData Xi for TS / the UE DNNC 224 sends an acknowledgement using uplink control plane signalling to the BS 210 (e.g., ACK). On receipt of the ACK from the UE 220, operations 1602i to 1652i are repeated for further input communication data for transmission in subsequent time slots from the BS 210 to UE 220.

[00250] In operation 1622, in the event that the BS DNNC 214 detects that the DL output communication signal would not satisfy the white interference noise level when transmitted as described with reference to FIGs. 1 to 15, then scrambling DL DNN communications is enabled and the BS 210 selects NNSI for reconfiguring the BS DL TX DNN 206 to generate DL output communication signal that satisfies the white interference noise level as described with reference to FIGs. 1 to 15. The selected NNSI includes the specific neural network layer(s) (e.g., NN layer #) that are scrambled, an initial seed (e.g., Seed), an identifier for a permutation random function (PRF), and a random permutation iteration number (e.g., RPermlt #) (e.g., NN layer #, Seed, PRF ID, RPermlt #). In operation 1622a/1623a, the BS DNNC 214 enables scrambling DL DNN communications and transmits using RRC control plane signalling over PDCCH the selected NNSI to the UE 220 (e.g, RRC Scramble DNN Enabled Req (NN layer#, Seed, PRF ID, RPermlt #)). In operation 1624b, when the UE DNNC 224 receives the RRC Scramble DNN Enabled Req message including the selected NNSI associated with the BS DL TX DNN 206, the UE DNNC 224 generates configuration data for reconfiguring the UE DL RX DNN 228 based on the NNSI and as described with reference to FIGs. 3a to 3c and/or FIGs. 4a to 4b, and sends a configuration instruction (e.g., Cfg(Scramble)) for scrambling configuration information for reconfiguring the UE DL RX DNN 228 of UE 220.

[00251] In operation 1623b, after the UE DL RX DNN 228 has been reconfigured, the UE DNNC 224 of the UE 220 uses uplink control plane signalling to send an RRC response message over the PUSCH indicating acknowledgement of the reconfiguration to the BS 210 (e.g., RRC Scramble DNN Enabled Resp (ACK)). On receipt of the acknowledgement, in operation 1624c, the BS DNNC 214 sends a configuration instruction (e.g., Cfg(Scramble)) with scrambling configuration information based on the selected NNSI to reconfigure the BS DL TX DNN structure 206.

[00252] In operation 1602j, input communication data for time slot j (e.g., l_Data Yj) is applied to the reconfigured BS DL TX DNN 206, which generates DL output communication signal corresponding to l_Data Yj for TS j (e.g., OC_Data Y). In operation 1604j, the BS DL TX DNN 206 outputs the OC_Data Yj to the BS DNNC 214 for transmission to the UE 220 as a transmit waveform signal over PDSCH in TS j. As described with reference to FIG. 5, on receiving OC_Data Yj, the BS DNNC 214 detects whether OC_Data Yj would satisfy a white interference noise level when transmitted, if this is the case, then the OC_Data Yj is, for example, buffered and transmitted over PDSCH in TS j as a transmission waveform signal. In operation 1616i, the BS DNNC 214 transmits the OC_Data Yj as a transmission waveform signal in PDSCH at TS j (e.g., PDSCH RF TX WAVEFORM (OC_DataYj, TS /)). The UE 220 receives the transmission signal waveform for TS j over PDSCH and processes (e.g., down conversion to base band) into received OC_Data Yj (e.g., Rx OC Yj)). In operation 1650j-1 , UE DNNC 224 applies or inputs the Rx OC Yj for TS j to the reconfigured UE RX DL DNN 228 for DNN processing. In operation 1650j-2, the UE RX DL DNN 228 processes the Rx OC Yj and generates reconstructed communication data for TS j corresponding to the l_Data Yj (e.g., RJData Yj). In operation 1652j, after the UE DL RX DNN 228 has successfully generated RJData Yj the UE DNNC 224 sends an acknowledgement using uplink control plane signalling over PUCCH to the BS 210 (e.g., ACK). On receipt of the ACK from the UE 220, operations 1602j to 1652j are repeated for further input communication data for transmission in subsequent time slots from BS 210 to UE 220 until it is detected that transmission of the DL output communication signal generated by BS DL TX DNN 206 for a subsequent time slot would not satisfy the white noise interference level.

[00253] In operation 1606a, the BS DNNC 214 detects that the current DL output communication signal for a current TS a would not satisfy the white interference noise level when transmitted over PDSCH as described with reference to FIGs. 1 to 15. the BS 210 selects an updated NNSI for reconfiguring the BS TX DL DNN 206 to generate DL output communication signal for TS a, and any subsequent time slots, that is forecast to satisfy the white interference noise level as described with reference to FIGs. 1 to 15. The selected updated NNSI includes data representative of at least a random permutation iteration number (e.g., RPermlt #) that is associated with generating the random permutation sequence for permuting the ordering of the current specific neural network layer(s) defined in operation 1623a. In operation 1614a, the BS DNNC 214 transmits using RRC control plane signalling over PDCCH with the selected updated NNSI and corresponding specific timing information (e.g., TS a, TS b, TS c, and TS d etc.) to the UE 220 (e.g., RRC I MAC Control message(RPermlt #, TS {a, b, c, d})). The specific timing information (e.g., TS {a, b, c, d}) describes when the UE DNNC 224 should apply the updated NNSI to the UE DL RX DNN 228. In operation 1614b, when the UE DNNC 224 receives the RRC I MAC Control message including the selected updated NNSI and specific timing information, the UE DNNC 224 stores the updated NNSI and corresponding specific timing information (e.g., TS {a, b, c, d}) in UE NNSI storage I look-up table I buffer 225a. In operation 1614b, the UE DNNC 224 sends using uplink control plane signalling over PUCCH an acknowledgement of receipt of the updated NNSI and specific timing information to the BS 210 (e.g., ACK). On receipt of the acknowledgement in operation 1614b, in operation 1610a, the BS DNNC 214 sends a configuration instruction (e.g., Cfg(RPermlt#)) with scrambling configuration information associated with the selected random permutation iteration number of the selected updated NNSI to reconfigure the BS DL TX DNN 206.

[00254] In operation 1602a, the BS 210 inputs or applies input communication data for time slot a (e.g., l_Data Za) to the reconfigured BS DL TX DNN 206, which generates DL output communication signal for TS a corresponding to l_Data Za (e.g., OC_Data Za). In operation 1604a, the OC_Data Za is provided to the BS DNNC 214 for transmission to the UE 220 over PDSCH as a transmit waveform signal in the specific time slot TS a. As described with reference to FIG. 5, on receiving OC_Data Za, the BS DNNC 214 detects whether transmission of OC_Data Za satisfies a white interference noise level, if this is the case, then the OC_Data Za is, for example, buffered and transmitted as a transmission waveform signal in the specific time slot TS a. If transmission of the OC_Data Za would not satisfy the white interference noise level, then operation 1606a is performed again.

[00255] In operation 1616a, the BS DNNC 214 transmits the OC_Data Za as a transmission waveform signal over PDSCH in TS a (e.g., PDSCH RF TX WAVEFORM (OC_DataZa, TS a)). The UE 220 receives the transmission signal waveform for TS a over the PDSCH and down converts into received OC_Data Za (e.g., Rx OC Za)). As described with reference to FIG. 5, the UE DNNC 224 may buffer the Rx OC Za for TS a until the UE DL RX DNN 228 is ready for processing TS a. Prior to processing TS a, in operation 1648a, the UE DNNC 224 retrieves the NNSI associated with TS a and generates configuration data for reconfiguring the UE DL RX DNN 228 based on the retrieved NNSI as described with reference to FIGs. 3a to 3c and/or FIGs. 4c to 4d, and sends a configuration instruction (e.g., Cfg(RPermit#)) with the scrambling configuration for reconfiguring the UE DL RX DNN 228. In operation 1650a-1 , the UE DNNC 224 applies Rx OC Za for TS a to the reconfigured UE DL RX DNN 228 for processing. In operation 1650a-2, the UE DL RX DNN 228 processes the Rx OC Za and generates reconstructed communication data for TS a corresponding to the l_Data Za (e.g., RJData Za). In operation 1652a, after the UE DL RX DNN 228 has successfully generated RJData Za the UE DNNC 224 sends an acknowledgement using uplink control plane signalling over PUCCH to the BS 210 (e.g., ACK). On receipt of the ACK from the UE 220, operations 1602a to 1652a are repeated for at least TS b, c and d and other subsequent time slots when the BS 210 transmits subsequent input communication data (e.g., l_Data Zb, l_Data Zc, l_Data Zd etc.) to UE 220. As illustrated, operations 1602a-1652a are repeated for TS d, where BS 210 transmits input communication data for TS d (e.g., l_Data Zd) as described in operations 1602d to 1652d.

[00256] During DL DNN communications between BS 210 and UE 220, the UE 220 and BS 210 are also performing UL DNN communications in which the UE 220 uses an UL transmitting DNN structure 226 (e.g., UE UL TX DNN) for transmissions on PUSCH to the BS 210, which receives and processes the transmissions using a UL receiving DNN structure 208 (e.g., BS UL RX DNN). It is assumed that the UE UL TX DNN 226 and BS UL RX DNN 208 have already been configured as described with reference to FIGs. 12 to 15. In operation 1606b, the UE DNNC 224 detects that a current UL output communication signal would not satisfy the white interference noise level when transmitted as described with reference to FIGs. 1 to 15. The UE 220 reconfigures its UE UL TX DNN 226. In this example, the UE 220 does not have the capabilities to select an updated NNSI, where instead, in operation 1607, the UE DNNC 224 using uplink control plane signalling over PUCCH requests communication resources (e.g., UL time slots/frequencies) for uplink scrambling transmission including updated NNSI (e.g., PLICCH Request (UL NNSI for Scrambling transmission)). On receipt of the request for communication resources and updated NNSI, the BS 210 selects an updated NNSI for reconfiguring the UE TX UL DNN 226 to generate UL output communication signals that would satisfy the white interference noise level for PUSCH as described with reference to FIGs. 1 to 15 (e.g., see FIG. 2a and FIG. 15 in operation 1508). The selected updated NNSI includes data representative of at least a random permutation iteration number (e.g., RPermlt #) that is associated with generating the random permutation sequence for permuting the ordering of the current specific neural network layer(s) of the UE UL TX DNN 226. In operation 1614e, the BS DNNC 214 transmits using DCI control plane signalling over PDCCH the selected updated NNSI and corresponding specific timing information for transmission (e.g., TS e) to the UE 220 (e.g., DCI Control message(RPermlt #, TS e)), where TS e indicates the time slot from when the updated NNSI should be applied. When the UE DNNC 224 receives the DCI message including the selected updated NNSI and specific timing information, TS e, the UE DNNC 224 stores the updated NNSI and corresponding specific timing information, TS e, in UE NNSI storage 225b. In operation 1614f, the UE DNNC 224 of the UE 220 sends, using uplink control plane signalling over PUCCH, an acknowledgement of receipt of the updated NNSI and specific timing information to the BS 210 (e.g., ACK).

[00257] In operation 1648b, for an UL transmission in TS e, the UE DNNC 224 sends a configuration instruction (e.g., Cfg(RPermlt#)) with scrambling configuration information based on random permutation iteration number of the selected updated NNSI for TS e to reconfigure the UE UL TX DNN 226. In operation 1602e, the reconfigured UE UL TX DNN 226 processes input communication data for TS e (e.g., l_Data Ze), which generates UL output communication signal for TS e corresponding to l_Data Ze (e.g., OC_Data Ze). In operation 1604e, the UE DNNC 224 of the UE 220 processes the OC_Data Ze for transmission as a transmit waveform signal in TS e over PUSCH to the BS 210. As described with reference to FIG. 5, on receiving OC_Data Ze, the UE DNNC 224 detects whether transmission of OC_Data Ze satisfies a white interference noise level, if this is the case, then the OC_Data Ze is, for example, buffered and transmitted as a transmission waveform signal in TS e over PLISCH to BS 210. Operation 1606b is repeated if transmission of the OC_Data Ze would not satisfy the white interference noise level.

[00258] In operation 1616e, the UE DNNC 224 transmits the OC_Data Ze as a transmission waveform signal in TS e over PLISCH to the BS 210 (e.g., PUSCH RF TX WAVEFORM (OC_DataZe, TS e)). The BE 210 receives the transmission signal waveform for TS e over PUSCH and processes (e.g., down conversion to base band) into received OC_Data Ze (e.g., Rx OC Ze)). As described with reference to FIG. 5, the BS DNNC 214 may buffer the Rx OC Ze for TS e until the BS UL RX DNN 208 is ready for processing TS e. Prior to processing TS e, in operation 1648e, the BS DNNC 214 retrieves the NNSI associated with TS e and generates configuration data for reconfiguring the BS UL RX DNN 208 based on the retrieved NNSI as described with reference to FIGs. 3a to 3c and/or FIGs. 4c to 4d, and sends a configuration instruction (e.g., Cfg(RPermit#)) with the scrambling configuration for reconfiguring the BS UL RX DNN 208. In operation 1650e-1 , the BS DNNC 214 applies Rx OC Ze for TS e to the reconfigured BS UL RX DNN 208 for DNN processing. In operation 1650e-2, the BS DL RX DNN 208 processes the Rx OC Ze and generates reconstructed communication data for TS e corresponding to the l_Data Ze (e.g., RJData Ze). In operation 1652e, after the BS DL RX DNN 208 successfully generates RJData Ze the BS DNNC 214 sends an acknowledgement using downlink control plane signalling over PDCCH to the UE 220 (e.g., ACK). On receipt of the ACK from the BS 210, the signal flow repeats operations 1602e to 1652e for subsequent time slots when the UE 220 has further input communication data for transmission to BS 210.

[00259] In operation 1627a, in this example, the BS 210 determines that scrambling DNN communications should be terminated (e.g., UE requested communications session to end, UE switches to RRCJDLE state or UE switches to RRCJNACTIVE state, or UE/BS connection failure, or communication session reverts to use conventional communications, etc.), in which case, the BS 210 uses RRC control plane signalling to indicate that scrambling DNN communications are to be disabled (e.g., RRC Scramble DNN Disable Req ( )). In operations 1628a, the BS DNNC 214 sends a configuration message to the corresponding BS TX/RX DNNs 206/208 for reverting to their original configuration for standard DNN communications, and/or releasing the associated DNN computing resources used for performing DNN communications for use in conventional communications and/or terminating the communications session. In operations 1628b, the UE DNNC 224 sends a configuration message to the corresponding UE TX/RX DNNs 226/228 for reverting to their original configuration for standard DNN communications, and/or releasing the associated DNN computing resources for performing DNN communications for use in conventional communications and/or terminating the communications session.

[00260] FIG. 17 shows a non-transitory computer-readable media 1700 according to some embodiments. The non-transitory computer-readable media 1700 may include a computer readable storage medium 1702 and/or input/output mechanism 1704 for enabling a computing system to access said computer-readable medium 1702. Although in this example the non-transitory computer-readable media 1700 is a USB stick, this is by way of example only and it is not so limited, the skilled person would appreciate the non-transitory computer-readable media 1700 may be any other type of computer-readable media or medium or computer-program product such as, for example, a CD, a DVD, a USB stick, a blue ray disk, flash drive etc. and/or any other computer-readable media or medium as the application demands. The non-transitory computer-readable media 1700 stores a computer program, computer program code and/or instructions, which when executed by one or more processors of an apparatus or system, causes the one or more processors of the apparatus or system to perform one or more of the methods, operations, processes of any signal-flow, flow-diagram, method and/or process as described herein, for example as disclosed in relation to the signal flow diagrams, flow diagrams and schematic diagrams of figures 1 to 16 and related features thereof.

[00261] Implementations of the methods or processes described herein may be realized as in digital electronic circuitry, integrated circuitry, specially designed ASICs (application specific integrated circuits), computer hardware, firmware, software, and/or combinations thereof. These may include computer program products (such as software stored on e.g., magnetic discs, optical disks, memory, Programmable Logic Devices) comprising computer readable instructions that, when executed by a processor, causes the processor to perform one or more of the methods and/or processes described herein. [00262] Any system feature as described herein may also be provided as a method or process feature, and vice versa. As used herein, means plus function features may be expressed alternatively in terms of their corresponding structure. In particular, method aspects may be applied to system aspects, and vice versa.

[00263] Furthermore, any, some and/or all features in one aspect can be applied to any, some and/or all features in any other aspect, in any appropriate combination. It should also be appreciated that particular combinations of the various features described and defined in any aspects of the invention can be implemented and/or supplied and/or used independently.

[00264] Although several embodiments have been shown and described, it would be appreciated by those skilled in the art that changes may be made in these embodiments without departing from the principles of this disclosure, the scope of which is defined in the claims and their equivalents.

Claims

WHAT IS CLAIMED IS:

1 . A method performed by a first device (104a) in communication with a second device (104b), the method comprising: processing (404) input communication data with a transmitting deep neural network, DNN, (106) for generating an output communication signal (118) for transmission to the second device; performing (406) a scrambling DNN operation in response to a forecasted transmission of the generated output communication signal not satisfying a white noise interference level, said scrambling DNN operation comprising: selecting (408) neural network scrambling information, NNSI, for reconfiguring the transmitting DNN to process the input communication data to generate a scrambled output communication signal, which when transmitted satisfies the white noise interference level; transmitting (414), to the second device, a control message comprising an indication of the NNSI and scrambling timing information for directing when the second device is to reconfigure a receiving DNN (109); and transmitting (416), to the second device, the scrambled output communication signal that satisfies the white noise interference level based on the scrambling timing information.

2. The method of claim 1 , wherein the NNSI comprises data representative of randomizing an ordering of inputs of the transmitting DNN, outputs of the transmitting DNN, and/or a set of neural network nodes of one or more neural network layers of the transmitting DNN for randomizing, and the performing the scrambling DNN operation further comprises: reconfiguring (410) the transmitting DNN based on the NNSI by randomizing an ordering of the inputs, outputs, or set of neural network nodes in the one or more neural network layers.

3. The method of any of claims 1 or 2, the performing the scrambling DNN operation further comprising: reconfiguring the transmitting DNN by performing, using one or more random permutation parameters, a random permutation of a set of neural network nodes in one or more neural network layers of the transmitting DNN of the first device, wherein the NNSI specifies data representative of the one or more random permutation parameters.

4. The method of claim 3, wherein the one or more random permutation parameters comprises one or more of: seed data, a random permutation iteration or order number, or identification of a seed generation and pseudo-randomizing function for use in performing the random permutation of the set of neural network nodes, and the performing the random permutation of the set of neural network nodes further comprises: generating, for each neural network layer, a random permutation sequence corresponding to the random permutation iteration or order number, wherein the random permutation sequence has a length equal to the number of neural network nodes in the set of neural network nodes of said each neural network layer; and randomizing the set of neural network nodes of said each neural network layer by applying the generated random permutation sequence to the set of neural network nodes.

5. The method of claim 4, wherein the seed data further comprises one or more of: an identity of the first device, an identity of the second device, an identity of a cell the second device is located, an identity of the cell the first device is located, timing slot information, frame identification number, and/or any other information associated with the first or second device, the method further comprising generating a seed, from the seed data, for performing the random permutation of the set of neural network nodes of the one or more neural network layers.

6. The method of any preceding claim, wherein the performing the scrambling DNN operation further comprises: reconfiguring the transmitting DNN based on performing a randomizing operation on the neural network nodes of a selected l-th neural network layer of the transmitting DNN, for 1<I<L and L is the number of neural network layers, wherein the NNSI specifies the selected l-th neural network layer of the transmitting DNN is to be randomized.

7. The method of claim 6, wherein selected l-th neural network layer comprises a set of N neural network nodes arranged in a specific ordering, where N>1, and the performing the randomizing operation on the selected l-th neural network layer of the transmitting DNN further comprises randomizing the set of N neural network nodes of the selected l-th neural network layer, based on: generating an /V-dimensional permutation matrix using a random permutation sequence of length N for randomizing the set of N neural network nodes of the selected l-th neural network layer; and randomizing the selected l-th neural network layer by multiplying the specific ordering of the set of N neural network nodes of the selected l-th neural network layer with the /V-dimensional permutation matrix to form a randomized ordering of the set of N neural network nodes.

8. The method of any preceding claim, wherein the selecting the NNSI further comprises: iteratively selecting NNSI based on: generating (672), for an i-th iteration i>0, an /-th randomization operation in relation to the inputs, outputs, or a set of neural network nodes of the one or more neural network layers of the transmitting DNN; reconfiguring (674) the transmitting DNN based on the i-th randomization operation; processing (676) the input communication data by inputting to the reconfigured transmitting DNN to generate a scrambled output communication signal; analyzing (678) the scrambled output communication signal to determine whether transmission of the scrambled output communication signal would satisfy the white noise interference level; responsive to the analysis indicating transmission of the scrambled output communication signal of the reconfigured transmitting DNN would satisfy the white noise interference level, indicating (682) the NNSI comprising the i-th randomization operation used to reconfigure the transmitting DNN; and responsive to the analysis indicating transmission of output communication signal of the reconfigured transmitting DNN would not satisfy the white noise interference level, updating (680) to a next i-th iteration and repeating the generating (672), the reconfiguring (674), the processing (676), and the analyzing (678) steps.

9. The method of claim 8, wherein the i-th randomization operation is a random permutation of an ordering of the set of neural network nodes, and further comprising iteratively selecting NNSI using the transmitting DNN of the first device to process the input communication data in each i-th random permutation iteration.

10. The method of any of claims 8 or 9, further comprising iteratively selecting NNSI by the first device based on a simulation of a transmission of the scrambled output communication signal over a simulated communication channel.

11 . The method of any preceding claim, further comprising storing any selected NNSI and corresponding whitening characteristics in a NNSI lookup table at the first device, and wherein the selecting the NNSI further comprises retrieving NNSI from the NNSI lookup table based on a likelihood the corresponding whitening characteristics indicate the selected NNSI will result in a transmission of the output communication signal of the transmitting DNN satisfying the white noise interference level.

12. The method of any preceding claim, further comprising: establishing a DNN connection with the second device for defining and configuring a transmitting DNN and receiving DNN for performing end-to-end communications therebetween; performing DNN communications with the second device for one or more time slots using the configured transmitting DNN and receiving DNN of the first device; and enabling performance of the scrambling DNN operation for one or more time slots according to the scrambling timing information when the output communication signal from the transmitting DNN does not satisfy the white noise interference level when transmitted.

13. The method of any preceding claim, the method further comprising, prior to performing the scrambling DNN operation, identifying whether an estimated spectral density of a forecast transmission of the output communication signal satisfies the white noise interference level.

14. The method of claim 13, wherein the identifying whether the estimated spectral density satisfies the white noise interference level further comprises: identifying that the estimated spectral density of the forecast transmission satisfies the white noise interference level when a power spectral density of the forecast transmission over a bandwidth of interest is below the white noise interference level or within a predetermined threshold region of the white noise interference level.

15. The method of claims 13 or 14, wherein the identifying whether the estimated spectral density satisfies the white noise interference level further comprises: identifying that the estimated spectral density of the forecast transmission does not satisfy the white noise interference level when it forms one or more spikes of interference above the white noise interference level.

16. The method of claims 13 or 14, wherein the estimated spectral density of the forecast transmission is estimated for each antenna output of the first device.

17. The method of any of claims 13 to 16, wherein the identifying whether the estimated spectral density of the forecast transmission satisfies the white noise interference level comprises: performing (406a) a spectral density estimation on the forecast transmission representing the output communication signal; and analysing (406b) the estimated spectral density in relation to a white noise power spectral density corresponding to the white noise interference level.

18. The method of claim 17, further comprising: comparing (406b) the estimated spectral density of the forecast transmission with a white noise spectral density associated with the white noise interference level; in response to the estimated spectral density of the forecast transmission being less then or substantially matching the white noise spectral density associated with the white noise interference level, indicating (406d) that the output communication signal satisfies the white noise interference level when transmitted; and in response to the estimated spectral density of the forecast transmission being greater than or substantially diverging from the white noise spectral density associated with the white noise interference level, indicating (406c) that the output communication signal does not satisfy the white noise interference level when transmitted.

19. The method as claimed any of any preceding claim, wherein the white noise interference level defines either: a constant amplitude for a flat white noise power spectral density over a bandwidth of interest; or a total power for a flat white noise power spectral density over a bandwidth of interest.

20. The method of any preceding claim, the method further comprising: receiving (1105a), from a third device (220b), a notification indicating that transmissions of output communication signals to the second device exceed an acceptable white noise interference level associated with the third device; adjusting (1105b) the white noise interference noise level to meet the acceptable white noise interference level; and the performing the scrambling DNN operation comprises analyzing whether the output communication signal satisfies the adjusted white noise interference level when transmitted.

21 . The method of any preceding claim, wherein the NNSI comprises data specifying one or more selected random permutation sequences used for randomizing one or more neural network layers of the transmitting DNN, wherein the transmitting the control message comprises sending a further control message specifying the selected random permutation sequences used.

22. The method of claim 21 , wherein the transmitting the control message comprises transmitting a selected random permutation iteration or order and seed information and corresponding time slot information using control plane signalling.

23. The method of any of claims 21 or 22, wherein the transmitting the control message comprises transmitting each control message as a Radio Resource Control, RRC, message, or a Medium Access Control, MAC, message, or a downlink control information, DCI, message.

24. The method of any preceding claim, further comprising transmitting further control messages from the first device to the second device using downlink control information, DCI, for use in indicating a selected random permutation sequence iteration or order used for reconfiguring the transmitting DNN according to the scrambling timing information.

25. A method performed by a second device in communication with a first device, the method comprising: receiving (432), from the first device, a control message indicating neural network scrambling information, NNSI, and scrambling timing information; receiving (442), from the first device, a communication signal transmitted according to the scrambling timing information; reconfiguring (448) a receiving deep neural network, DNN, of the second device according to the received NNSI and the scrambling timing information; processing (450) the received communication signal with the receiving DNN for generating reconstructed communication data represented by the received communication signal; and sending (452) the reconstructed communication data to a data sink of the second device or sending the reconstructed communication data to one or more upper protocol layers of a protocol stack of the second device.

26. The method of claim 25, the receiving (432) the control message further comprising: receiving (432), from the first device, one or more control messages, each control message indicating NNSI and corresponding scrambling timing information; and storing (434) the received NNSI in storage for use in relation to the scrambling timing information.

27. The method of claim 26, wherein the scrambling timing information comprises one or more time slots for receiving transmissions from the first device, the method further comprising: the receiving (442), from the first device, the communication signal further comprising receiving, from the first device, the communication signal in a specific time slot of the one or more time slots; the reconfiguring (448) the receiving DNN of the second device using the NNSI retrieved from storage; and processing (450) the received communication signal with the receiving DNN to produce reconstructed communication data represented by the received communication signal for the specific time slot.

28. The method of any of claims 25 to 27, wherein the NNSI comprises data representative of randomizing an ordering of inputs, outputs, or a set of neural network nodes of one or more neural network layers of a transmitting DNN of the first device, and the reconfiguring the receiving DNN of the second device further comprising reconfiguring the receiving DNN of the second device according to the scrambling timing information using the NNSI to reverse the randomization applied to the inputs, outputs, and/or neural network layers of the transmitting DNN.

29. The method of claim 28, wherein the NNSI specifies a random permutation of the ordering of a set of neural network nodes in the one or more neural network layers of the transmitting DNN.

30. The method of claim 29, wherein the NNSI comprises one or more random permutation parameters comprising: seed data, a random permutation iteration or order number, or identification of a seed generation and pseudo-randomizing function for use in performing the random permutation of the set of neural network nodes of one or more neural network layers, and the reconfiguring the receiving DNN of the second device comprises: generating, for each neural network layer of the receiving DNN corresponding to each neural network layer of the transmitting DNN, an inverse random permutation sequence corresponding to the random permutation iteration or order number, wherein the inverse random permutation sequence has a length equal to the number of neural network nodes in the set of neural network nodes of said each neural network layer of the receiving DNN, and derandomizing, for each neural network layer of the receiving DNN corresponding to each neural network layer of the transmitting DNN, the set of neural network nodes of said each neural network layer of the receiving DNN by applying the generated inverse random permutation sequence to the set of neural network nodes.

31 . The method of claim 30, wherein the seed data further comprises one or more of: an identity of the first device, an identity of the second device, an identity of a cell the second device is located, an identity of the cell the first device is located, timing slot information, frame identification number, and/or any other information associated with the first or second device, the method further comprising generating a seed, from the seed data, for performing an inverse random permutation of the set of neural network nodes of the neural network layers of the receiving DNN corresponding to each neural network layer of the transmitting DNN.

32. The method of any of claims 25 to 31 , wherein the NNSI specifies an 1-th neural network layer of the transmitting DNN is randomized, for 1<I<L and L is the number of neural network layers of the transmitting DNN including an input layer and an output layer, and reconfiguring the receiving DNN further comprising performing an inverse randomizing operation on the set of neural network nodes of an (L-l+1)-th neural network layer of the receiving DNN.

33. The method of claim 32, wherein the (L-l+1)-th neural network layer of the receiving DNN has a set of N neural network nodes, N>1 , with a specific ordering, and performing the inverse randomizing operation on the set of neural network nodes of (L-l+1)-th neural network layer of the receiving DNN comprising: generating a A/-dimensional permutation matrix based on a random permutation for the (L-l+1)-th neural network layer, N>1 inverting the /V-dimensional permutation matrix to generate an inverted N- dimensional permutation matrix; and multiplying the specific ordering of the set of /V-neural network nodes of the (L- l+1)-th neural network layer with the inverted /V-dimensional permutation matrix to form an inverse randomized ordering of the set of N neural network nodes.

34. The method of any of claims 25 to 33, wherein the NNSI specifies a random permutation iteration or order, seed information, and corresponding time slot information and the control message is received using control plane signalling, wherein the reconfiguring the receiving DNN of the second device further comprising reconfiguring the receiving DNN of the second device using the random permutation iteration or order and seed for processing the received communication signal transmitted from the first device according to the time slot information.

35. The method of claim 34, wherein the control message is a Radio Resource Control, RRC, message, or a Medium Access Control, MAC, message, or a downlink control information, DCI, message.

36. The method of any of claims 25 to 35, the method further comprising receiving, from the first device, one or more further control messages using downlink control information, DCI, for use in indicating a selected random permutation sequence iteration or order and scrambling timing information for reconfiguring the receiving DNN according to the scrambling timing information.

37. The method of any preceding claim, wherein the first device (104a) is a base station (210), and the second device (104b) is a user equipment (220).

38. The method of any of claims 1 to 36, wherein the first device (104a) is a user equipment (220), and the second device (104b) is a base station (210).

39. A computer-readable medium (1700) comprising computer readable instructions stored thereon that, when executed by a computer, cause the computer to perform the method of any preceding claim.

40. A first device (210) comprising one or more processors (212) and a memory (213), the memory storing computer readable instructions that, when executed by the one or more processors, cause the first device (210) to perform the method of any of claims 1 to 24 and 37 to 38.

41 . A second device (220) comprising one or more processors (222) and a memory (223), the memory (223) storing computer readable instructions that, when executed by the one or more processors (222), cause the second device (210) to perform the method of any of claims 25 to 38.

42. An apparatus (210, 220) comprising: one or more antennas (203a/b); one or more processors (212, 222); and a memory (213, 223), wherein the one or more processors (212, 222) are connected to the memory (213, 223) and the one or more antennas (203a/b), and the memory (213, 223) further storing computer readable instructions that, when executed by the one or more processors (212, 222), cause the apparatus (210, 220) to perform the method of any of claims 1 to 38.

43. A communication system (200) comprising: a first device (210) configured according to claim 40; and a second device (220) configured according to claim 41 ; wherein the first device (210) and the second device (220) establish a deep neural network communication session.