CN114143147A - Unauthorized large-scale Internet of things equipment detection method based on deep learning - Google Patents
Unauthorized large-scale Internet of things equipment detection method based on deep learning Download PDFInfo
- Publication number
- CN114143147A CN114143147A CN202111333654.2A CN202111333654A CN114143147A CN 114143147 A CN114143147 A CN 114143147A CN 202111333654 A CN202111333654 A CN 202111333654A CN 114143147 A CN114143147 A CN 114143147A
- Authority
- CN
- China
- Prior art keywords
- internet
- things
- things equipment
- active
- neural network
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L25/00—Baseband systems
- H04L25/02—Details ; arrangements for supplying electrical power along data transmission lines
- H04L25/0202—Channel estimation
- H04L25/0224—Channel estimation using sounding signals
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L25/00—Baseband systems
- H04L25/02—Details ; arrangements for supplying electrical power along data transmission lines
- H04L25/0202—Channel estimation
- H04L25/024—Channel estimation channel estimation algorithms
- H04L25/0254—Channel estimation channel estimation algorithms using neural network algorithms
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L67/00—Network arrangements or protocols for supporting network services or applications
- H04L67/01—Protocols
- H04L67/12—Protocols specially adapted for proprietary or special-purpose networking environments, e.g. medical networks, sensor networks, networks in vehicles or remote metering networks
Abstract
The invention discloses an unauthorized large-scale Internet of things equipment detection method based on deep learning, which is carried out according to the following steps: 1) inserting a pilot frequency sequence in a data frame sent by active Internet of things equipment; 2) the active Internet of things equipment simultaneously transmits the data of the active Internet of things equipment to an access point on the same base station; 3) separating a pilot frequency sequence and a data sequence of received signals of a plurality of pieces of Internet of things equipment, detecting active users by using a pilot frequency sequence matrix known by a base station, and constructing a deep neural network input vector; 4) obtaining a training sample set required by the deep neural network by using a known algorithm; 5) constructing a deep neural network model and training the deep neural network model; 6) and recovering the data of the plurality of active Internet of things devices by adopting the trained deep neural network model.
Description
Technical Field
The invention relates to the technical field of Internet of things multi-device detection, in particular to an unauthorized Internet of things large-scale device detection method based on deep learning.
Background
In large-scale internet of things networks, data streams are typically sparse and low-rate, and limited resource blocks and thousands of internet of things devices are served by the same access point/base station, so it is not efficient to allocate one resource block for each internet of things device. To solve this problem, an unauthorized scheme in a large-scale internet of things network has attracted great attention. On the other hand, machine learning algorithms have evolved greatly over the past few years, and the potential in wireless communication for the internet of things has been recognized. In addition, machine learning algorithms, in particular deep learning algorithms, have been proposed in modulation, constellation design and resource allocation in wireless networks.
The traditional way to separate signals is to use Serial Interference Cancellation (SIC) at the receiving end to achieve separation of non-orthogonal superimposed signals. This approach requires the implementation of additional channel estimation algorithms, and is therefore highly complex and subject to error propagation.
Disclosure of Invention
The invention aims to solve the technical problem of providing an unauthorized large-scale Internet of things equipment detection method based on deep learning, which does not need a channel estimation algorithm, can utilize a pilot sequence multi-user detection to recover data, and is superior to the traditional SIC detector.
The invention adopts the following technical scheme for solving the technical problems:
an unauthorized large-scale Internet of things equipment detection method based on deep learning is characterized in that an unauthorized large-scale Internet of things environment comprises an access point and L Internet of things equipment, k Internet of things equipment in an active state (namely active Internet of things equipment) in the L Internet of things equipment are in a sleep state, and k is less than or equal to L. The method specifically comprises the following steps:
respectively inserting different pilot frequency sequences into a data symbol sent by each active Internet of things device to obtain a total frame of each active Internet of things device;
step two, the k active Internet of things devices simultaneously transmit the total frames to an access point on the base station;
separating a pilot frequency sequence and a data sequence of the signals of the k active Internet of things devices received by the access point, detecting the active Internet of things devices by using a pilot frequency sequence matrix at the base station, and constructing an input vector;
step four, constructing a deep neural network model and a training sample set required by training the deep neural network model, and training the deep neural network model by using the training sample set;
and step five, recovering the data symbols sent by the k active Internet of things devices by adopting the trained deep neural network model.
Compared with the prior art, the invention adopting the technical scheme has the following technical effects:
(1) the method can detect symbols of any number of devices in unauthorized communication, and the model of the method can be used as an online detector in an unauthorized large-scale Internet of things network after offline training.
(2) The method supports unauthorized communication and does not introduce additional signaling overhead.
(3) The method performs multi-user detection based on the pilot signal so that no additional channel estimation algorithm is required.
Drawings
Fig. 1 is a system model diagram of a large-scale internet of things environment link scenario.
Fig. 2 is a flowchart of an unauthorized large-scale internet of things device detection method based on deep learning.
FIG. 3 is a diagram of a deep neural network model according to the present invention.
Detailed Description
For a further understanding of the invention, reference should be made to the following detailed description taken in conjunction with the accompanying drawings. It is to be understood that the examples are illustrative of the invention and not limiting.
The embodiment of the invention provides an unauthorized large-scale Internet of things equipment detection method based on deep learning, which does not need a channel estimation algorithm, can perform multi-user detection by using a pilot frequency sequence and recover data, and is superior to a traditional SIC detector.
As shown in fig. 1, the unauthorized large-scale internet of things environment of this example includes an access point and L internet of things devices, where k internet of things devices in an active state are among the L internet of things devices, and other internet of things devices are in a sleep state.
As shown in fig. 2, an embodiment of the present invention provides a flowchart of an unauthorized large-scale internet of things device detection method based on deep learning, where the method includes the following steps:
respectively inserting different pilot frequency sequences into a data symbol sent by each active Internet of things device to obtain a total frame of each active Internet of things device;
step two, the k active Internet of things devices simultaneously transmit the total frames to an access point on the base station;
separating a pilot frequency sequence and a data sequence of the signals of the k active Internet of things devices received by the access point, detecting the active Internet of things devices by using a pilot frequency sequence matrix at the base station, and constructing an input vector;
step four, constructing a deep neural network model and a training sample set required by training the deep neural network model, and training the deep neural network model by using the training sample set;
and step five, recovering the data symbols sent by the k active Internet of things devices by adopting the trained deep neural network model, and outputting estimated data symbols.
Specifically, in the step one, in order to distinguish different active internet of things devices, a pilot sequence needs to be inserted into a data symbol sent by the active internet of things devices, each active internet of things device uses a different pilot sequence, and a total frame of the ith internet of things device after the pilot sequence is inserted into the ith internet of things device can be represented as:
wherein the content of the first and second substances,and xi,d=[xi,d(Ns+1),...,xi,d(Ns+Nd)]Are respectively the ithPilot sequence and data symbol of internet of things device, 1,2s,Ns+1,...,Ns+NdN representing a total frame of an ith active Internet of things devices+NdOne time slot, NsAnd NdRespectively, the length of the pilot sequence and the data symbol length. When the quantity of the IOT equipment is less, NsThe pilot sequences of different internet of things devices are orthogonal, and active devices can be completely identified at the moment; when the number of IOT devices is large, consider the overload system, i.e. Ns< L, matrix analysis needs to be performed at the base station.
Specifically, in the step two, since the access mode is unauthorized communication, the active internet of things devices may directly transmit their data to the access point without applying for the data from the base station. Assuming that data is transmitted frame by frame, in the pilot sequence segment, the signal received by the access point can be expressed as:
in the data section, the signal received by the access point can be represented as:
wherein y isp=[y(1),y(2),...y(Ns)],yd=[y(Ns+1),y(Ns+2),...y(Ns+Nd)],aiE {0,1} represents the activity of the ith active internet of things device. PiAnd hiIs the transmission power and channel coefficient of the ith active internet of things device, N ═ N (1), N (2),. N (N)s+Nd)]Is additive white gaussian noise.
Specifically, in step three, the base station obtains B ═ B according to the product of the pseudo-inverse matrix of the pilot sequence matrix S and the pilot sequence segment signal received by the access point1,b2,...,bL]=S+ypIn which S is+Is a pilot sequence matrixPseudo-inverse matrix of S, S ═ S1,s2,...,sL],For the pilot sequence of the ith internet of things device, L is 1,2, …, L.
B is tolAnd a threshold value a0Comparing if greater than a0Then the activity of the ith internet of things deviceOtherwiseObtaining a sequence of L internet of things equipment activities obtained by calculation at a base station
Data segment signal y received by access pointdThe method is divided into real data and virtual data. Furthermore, in an LSTM network, the input and output data sizes should be the same, and N is estimated in the model under considerationdData symbols of length, whereby the input data has a length of Nd. On the basis, the active signal of the Internet of things equipment is modified, and the active signal is input (namely N) by adding the received active signal of the same Internet of things equipmentdDimension vector yi,p) Extension to NdLength, so the modified deep neural network input vector is the input of 2 inputs (real and imaginary parts) of the data signal and k active signals, i.e.
Specifically, an activity sequence a ═ a of L pieces of internet-of-things equipment is constructed1,a2,...,aL]In the step a, k elements are randomly set as 1, and the rest are 0 (namely k active internet of things devices, and the rest of the internet of things devices are in a sleep state). Then randomly generating N for the ith active Internet of things equipmentd×log2Mi(bit) data, generating x by M-ary modulationi,dConstructing an input vectorAnd constructing all 0 input vectors with the length same as that of the input vectors of the active Internet of things equipment for the Internet of things equipment in a sleep state. And forming a training sample set by using the input vectors of the L pieces of Internet of things equipment.
Specifically, as shown in fig. 3, the deep neural network model in step five is divided into three parts, where the first part is an input layer, and the input layer is used for receiving the processed signal; the second part is a hidden layer, and an LSTM neural unit is adopted for realizing function fitting of a detection algorithm; and the third part is an output layer which is a full connection layer and generates an estimation data symbol for each piece of equipment of the Internet of things. The model outputs data asThe accuracy of the training model is measured by calculating a loss function, and the mathematical expression iswmAnd for the mth neural network parameter, training the deep neural network model by taking the minimum loss function as a criterion, wherein the optimization process mainly adopts an Adam optimization algorithm, and continuously iterates and updates the numerical value of the hyperparameter according to the solved optimization vector until an optimal solution is found. Wherein, the error rate of the information is continuously counted in the training process, and the standard of the completion of the training isMax { P ] under the condition that the loss function is kept in a stable statel(e)}<P,Pl(e) Is the error probability of the ith internet of things device.
Specifically, in the sixth step, the data processed in the fourth step is sent to an input layer, and an estimated data symbol of each internet of things device is output. For a signal composed of L data sent by Internet of things equipment, the input is L +2 (2 inputs (real part and imaginary part) of a data signal and L inputs of active signals) and the length is NdThe output layer is divided into L groups, and the ith group outputs the estimated data symbol of the ith Internet of things equipmentIs expressed as length Nd。
The present invention and its embodiments have been described above schematically, without limitation, and what is shown in the drawings is only one of the embodiments of the present invention, and the actual structure is not limited thereto. Therefore, if the person skilled in the art receives the teaching, without departing from the spirit of the invention, the person skilled in the art shall not inventively design the similar structural modes and embodiments to the technical solution, but shall fall within the scope of the invention.
Claims (8)
1. The unauthorized large-scale Internet of things equipment detection method based on deep learning is characterized in that the unauthorized large-scale Internet of things environment is provided with an access point and L Internet of things equipment, k active Internet of things equipment is arranged in the L Internet of things equipment, k is less than or equal to L, and L is a positive integer; the method specifically comprises the following steps:
respectively inserting different pilot frequency sequences into a data symbol sent by each active Internet of things device to obtain a total frame of each active Internet of things device;
step two, the k active Internet of things devices simultaneously transmit the total frames to an access point on the base station;
separating a pilot frequency sequence and a data sequence of the signals of the k active Internet of things devices received by the access point, detecting the active Internet of things devices by using a pilot frequency sequence matrix at the base station, and constructing an input vector;
step four, constructing a deep neural network model and a training sample set required by training the deep neural network model, and training the deep neural network model by using the training sample set;
and step five, recovering the data symbols sent by the k active Internet of things devices by adopting the trained deep neural network model.
2. The deep learning-based unauthorized large-scale internet of things device detection method according to claim 1, wherein in the first step, the total frame of the i-th active internet of things device is represented as:
wherein the content of the first and second substances,and xi,d=[xi,d(Ns+1),...,xi,d(Ns+Nd)]Pilot sequence and data symbol, 1, 2.., N, of the ith active internet of things device, respectivelys,Ns+1,...,Ns+NdN representing a total frame of an ith active Internet of things devices+NdOne time slot, NsAnd NdRespectively, the length of the pilot sequence and the length of the data symbol.
3. The deep learning-based unauthorized large-scale internet of things equipment detection method according to claim 1, wherein in step two, the pilot sequence segment signal received by the access point is represented as:
the data segment signal received by the access point is represented as:
wherein y isp=[y(1),y(2),...y(Ns)],yd=[y(Ns+1),y(Ns+2),...y(Ns+Nd)],aiE {0,1} represents the activity of the ith active Internet of things device, PiAnd hiIs the transmission power and channel coefficient of the ith active internet of things device, N ═ N (1), N (2),. N (N)s+Nd)]Is additive white gaussian noise.
4. The deep learning-based unauthorized large-scale internet of things equipment detection method according to claim 3, wherein the third step is specifically:
(1) the base station multiplies the pseudo inverse matrix of the pilot sequence matrix S by the pilot sequence segment signal received by the access point to obtain B ═ B1,b2,...,bL]=S+ypIn which S is+Is a pseudo-inverse of the pilot sequence matrix S, S ═ S1,s2,...,sL],A pilot sequence of the ith internet of things device, where L is 1,2, …, L;
(2) b is tolAnd a threshold value a0Comparing if greater than a0Then the activity of the ith internet of things deviceOtherwiseObtaining a sequence of L internet of things equipment activities obtained by calculation at a base station
(3) Data segment signal y received by access pointdDivided into real dataAnd dummy dataTwo parts are as follows:
5. The deep learning-based unauthorized large-scale internet of things equipment detection method according to claim 4, wherein the fourth step is specifically:
(1) constructing an activity sequence A ═ a of L pieces of Internet of things equipment1,a2,...,aL]Randomly setting k elements as 1 and the rest as 0 in the A; the Internet of things equipment corresponding to the element with the median value of 1 in the activity sequence is active Internet of things equipment;
(2) randomly generating N for the ith active Internet of things equipmentd×log2MiData symbols of the data are modulated by M elements to generate xi,dConstructing an input vector according to the method of the first step to the third step;
(3) constructing all 0 input vectors with the same length as the input vectors in the step (2) for the rest Internet of things devices except the active Internet of things devices;
(4) and (4) forming a training sample set by the input vectors constructed in the steps (2) and (3).
6. The deep learning-based unauthorized large-scale internet of things device detection method according to claim 1, wherein the deep neural network model in step four is divided into three parts, the first part is an input layer, the second part is a hidden layer using an LSTM neural unit, the third part is an output layer using a fully-connected layer, and the output layer outputs the estimated data symbols of each internet of things device.
7. The deep learning-based unauthorized large-scale internet of things equipment detection method according to claim 6, wherein the output layers of the deep neural network model are divided into L groups, and the I group outputs the estimated data symbol of the I-th internet of things equipment; the mathematical expression of the loss function of the deep neural network model iswmFor the mth neural network parameter, λ is the regularization parameter.
8. The deep learning-based unauthorized large-scale Internet of things equipment detection method according to claim 7, characterized in that training of a deep neural network model is performed by using an Adam optimization algorithm to find an optimal solution on the basis of the minimum loss function; wherein, the error rate of the information is continuously counted in the training process, and the standard of the training completion is that the error rate meets max { P under the condition that the loss function keeps a stable statel(e) P is greater than P, P is an error probability threshold, Pl(e) Is the error probability of the ith internet of things device.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111333654.2A CN114143147A (en) | 2021-11-11 | 2021-11-11 | Unauthorized large-scale Internet of things equipment detection method based on deep learning |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111333654.2A CN114143147A (en) | 2021-11-11 | 2021-11-11 | Unauthorized large-scale Internet of things equipment detection method based on deep learning |
Publications (1)
Publication Number | Publication Date |
---|---|
CN114143147A true CN114143147A (en) | 2022-03-04 |
Family
ID=80392947
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202111333654.2A Pending CN114143147A (en) | 2021-11-11 | 2021-11-11 | Unauthorized large-scale Internet of things equipment detection method based on deep learning |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN114143147A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114915361A (en) * | 2022-05-13 | 2022-08-16 | 西安交通大学 | Internet of things uplink signal detection method based on small sample learning |
-
2021
- 2021-11-11 CN CN202111333654.2A patent/CN114143147A/en active Pending
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114915361A (en) * | 2022-05-13 | 2022-08-16 | 西安交通大学 | Internet of things uplink signal detection method based on small sample learning |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Wu et al. | A CNN-based end-to-end learning framework toward intelligent communication systems | |
Hasan et al. | The role of deep learning in NOMA for 5G and beyond communications | |
Emir et al. | DeepMuD: Multi-user detection for uplink grant-free NOMA IoT networks via deep learning | |
Zhang et al. | Deep learning aided OFDM receiver for underwater acoustic communications | |
CN113708855B (en) | OTFS data driving and receiving method, system and medium based on deep learning | |
CN112637094A (en) | Multi-user MIMO receiving method based on model-driven deep learning | |
CN112215335B (en) | System detection method based on deep learning | |
CN110311876A (en) | The implementation method of underwater sound OFDM receiver based on deep neural network | |
CN111865845A (en) | MIMO user detection and channel estimation device and method | |
CN108206795B (en) | Blind authentication method and system of frequency selective fading channel based on confidence transfer | |
CN114143147A (en) | Unauthorized large-scale Internet of things equipment detection method based on deep learning | |
Van Huynh et al. | Transfer learning for signal detection in wireless networks | |
Ye et al. | Bilinear convolutional auto-encoder based pilot-free end-to-end communication systems | |
Zhao et al. | Multi-task learning based underwater acoustic OFDM communications | |
Liu et al. | OFDM-Based Digital Semantic Communication with Importance Awareness | |
CN110944002B (en) | Physical layer authentication method based on exponential average data enhancement | |
Chen et al. | Blind nonlinear system identification based on a constrained hybrid genetic algorithm | |
Bhatt et al. | Analysis of the fifth generation NOMA system using LSTM algorithm | |
CN113242201B (en) | Wireless signal enhanced demodulation method and system based on generation classification network | |
CN113037409B (en) | Large-scale MIMO system signal detection method based on deep learning | |
Ivanov et al. | Hybrid noise-resilient deep learning architecture for modulation classification in cognitive radio networks | |
CN111769975A (en) | MIMO system signal detection method and system | |
Ma et al. | Design of DBN based Demodulator in Underwater Wireless Optical Communications | |
Ahmad et al. | Wavelet-based massive MIMO-NOMA with advanced channel estimation and detection powered by deep learning | |
McCaskey et al. | Machine learning applied to an RF communication channel |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination |