CN112867010B - Radio frequency fingerprint embedded real-time identification method and system based on convolutional neural network - Google Patents
Radio frequency fingerprint embedded real-time identification method and system based on convolutional neural network Download PDFInfo
- Publication number
- CN112867010B CN112867010B CN202110048449.5A CN202110048449A CN112867010B CN 112867010 B CN112867010 B CN 112867010B CN 202110048449 A CN202110048449 A CN 202110048449A CN 112867010 B CN112867010 B CN 112867010B
- Authority
- CN
- China
- Prior art keywords
- radio frequency
- frequency fingerprint
- fingerprint identification
- signal
- time
- 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.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 38
- 238000013527 convolutional neural network Methods 0.000 title claims abstract description 18
- 238000004891 communication Methods 0.000 claims abstract description 21
- 238000007781 pre-processing Methods 0.000 claims abstract description 21
- 238000012544 monitoring process Methods 0.000 claims abstract description 19
- 238000012795 verification Methods 0.000 claims abstract description 8
- 238000012545 processing Methods 0.000 claims description 27
- 238000012549 training Methods 0.000 claims description 22
- 230000006870 function Effects 0.000 claims description 16
- 230000008569 process Effects 0.000 claims description 11
- 238000005070 sampling Methods 0.000 claims description 10
- 238000003860 storage Methods 0.000 claims description 9
- 230000004913 activation Effects 0.000 claims description 8
- 238000011176 pooling Methods 0.000 claims description 8
- 238000012360 testing method Methods 0.000 claims description 8
- 238000001914 filtration Methods 0.000 claims description 7
- 238000006243 chemical reaction Methods 0.000 claims description 5
- 230000011218 segmentation Effects 0.000 claims description 5
- 238000010606 normalization Methods 0.000 claims description 4
- 230000008859 change Effects 0.000 claims description 3
- 230000009467 reduction Effects 0.000 claims description 3
- 238000013528 artificial neural network Methods 0.000 abstract description 12
- 238000011161 development Methods 0.000 abstract description 9
- 238000000605 extraction Methods 0.000 description 9
- 230000018109 developmental process Effects 0.000 description 8
- 238000010586 diagram Methods 0.000 description 8
- 230000006872 improvement Effects 0.000 description 7
- 238000013135 deep learning Methods 0.000 description 6
- 238000005516 engineering process Methods 0.000 description 5
- 238000004364 calculation method Methods 0.000 description 4
- QVFWZNCVPCJQOP-UHFFFAOYSA-N chloralodol Chemical compound CC(O)(C)CC(C)OC(O)C(Cl)(Cl)Cl QVFWZNCVPCJQOP-UHFFFAOYSA-N 0.000 description 2
- 230000007547 defect Effects 0.000 description 2
- 208000009115 Anorectal Malformations Diseases 0.000 description 1
- 235000011034 Rubus glaucus Nutrition 0.000 description 1
- 244000235659 Rubus idaeus Species 0.000 description 1
- 235000009122 Rubus idaeus Nutrition 0.000 description 1
- 238000013500 data storage Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000013508 migration Methods 0.000 description 1
- 230000005012 migration Effects 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 238000002360 preparation method Methods 0.000 description 1
- 230000004044 response Effects 0.000 description 1
Images
Classifications
-
- 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
-
- 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
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D30/00—Reducing energy consumption in communication networks
- Y02D30/70—Reducing energy consumption in communication networks in wireless communication networks
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Molecular Biology (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Measurement Of The Respiration, Hearing Ability, Form, And Blood Characteristics Of Living Organisms (AREA)
- Mobile Radio Communication Systems (AREA)
Abstract
The invention relates to computer software and the technical field of wireless communication safety, in particular to a radio frequency fingerprint embedded real-time identification method and a radio frequency fingerprint embedded real-time identification system based on a convolutional neural network, wherein the method comprises the following steps: collecting and storing wireless signals received by an antenna; when a new wireless signal is found, reading signal data in sequence according to a preset acquisition point number and a preset time length; preprocessing the read signal data; inputting the preprocessed signal data into a pre-established radio frequency fingerprint identification model, and outputting a radio frequency equipment label corresponding to the signal data to obtain a radio frequency fingerprint identification result. The system provided by the invention has mobility and real-time performance, guarantees the wireless communication safety of each mobile device, meets the wireless communication identity verification requirements of the mobile devices, and realizes the real-time identification of the radio frequency fingerprint of a wireless signal by using an embedded development board with low cost and low power consumption, carrying SDR (software defined device), a light weight neural network and a real-time monitoring module.
Description
Technical Field
The invention relates to computer software and the technical field of wireless communication safety, in particular to a radio frequency fingerprint embedded real-time identification method and a radio frequency fingerprint embedded real-time identification system based on a convolutional neural network.
Background
The radio frequency fingerprint means that small random flaws still exist on a radio frequency circuit on the premise that the product is qualified in the process of manufacturing the wireless communication equipment. The flaws are unique and can be kept unchanged for a long time, and finally, the flaws are combined with various factors such as frequency offset and phase noise of each discrete device of a transmitter, nonlinearity of a local vibration source, filter characteristics and the like to form a radio frequency fingerprint with individual characteristics, and the fingerprint similar to a biological fingerprint can uniquely identify a certain individual. Because the radio frequency fingerprint is based on the characteristics of the radio frequency circuit of the equipment, and has uniqueness and invariance, the identity authentication of the wireless communication equipment can be effectively realized through the radio frequency fingerprint.
The radio frequency fingerprint identification technology can identify a specific transmitter through the characteristics of a physical layer of a transmitting terminal, can perform safety protection on various wireless terminals in different scenes, and has huge application prospects in wireless communication networks in the civil and military fields.
The traditional radio frequency fingerprint identification method is based on manual selection of features, and has the defects of poor method universality, low identification rate and the like, so that the radio frequency fingerprint identification system is difficult to popularize. With the development of big data and deep learning technology, the radio frequency fingerprint identification method based on deep learning surpasses the traditional method based on artificial characteristics in the aspects of method universality and identification accuracy. At present, a radio frequency fingerprint identification system combining a constellation diagram and deep learning is researched and proposed, but the deep learning method based on images has the problems of numerous neural network parameters, large calculated amount and the like, and the radio frequency fingerprint identification of real-time signals needs a lot of time, and is slow in response speed, low in instantaneity and low in identification efficiency. With the development of the internet of things, the mobile property of wireless communication is more and more common, meanwhile, in the prior art, only fixed-point radio frequency fingerprint identification such as a base station is concerned, the requirement of large computation amount of an image processing neural network on hardware is higher, and the cost and the mobility in the prior art are higher.
Because the prior art mostly uses a signal constellation diagram as the input of a neural network, originally only two-dimensional signals are converted into three-dimensional images for recognition, although the characteristics of the signals are extracted, the parameters and the operation amount of the network are increased, and the recognition speed is limited. Meanwhile, no technology for realizing unmanned real-time monitoring of wireless signals can be realized, so that no technology has the function of real-time radio frequency fingerprint identification at present.
At present, mainstream radio frequency fingerprint feature extraction is divided into manual extraction and deep learning extraction, the manual extraction method is weak in universality of signals, deep learning is achieved through feature extraction of three-dimensional images, calculated amount is increased, certain requirements are placed on hardware, and therefore the system is low in identification speed, high in system cost and low in mobility. And the functions of unmanned real-time signal monitoring and signal radio frequency fingerprint identification are not realized in the current mainstream technology.
Disclosure of Invention
The invention aims to overcome the defects of the prior art and provides a radio frequency fingerprint embedded real-time identification method and system based on a convolutional neural network.
In order to achieve the above object, the present invention provides a radio frequency fingerprint embedded real-time identification method based on a convolutional neural network, the method comprising:
collecting and storing wireless signals received by an antenna;
when a new wireless signal is found, reading signal data in sequence according to a preset acquisition point number and a preset time length;
preprocessing the read signal data;
and inputting the preprocessed signal data into a pre-established radio frequency fingerprint identification model, and outputting a radio frequency equipment label corresponding to the signal data to obtain a radio frequency fingerprint identification result.
As an improvement of the above method, the wireless signals received by the antenna are collected and stored; the method specifically comprises the following steps:
and acquiring wireless signals received by the antenna, processing the wireless signals, outputting sampling signals, performing low-pass filtering processing, and storing the sampling signals.
As an improvement of the above method, the signal data read is preprocessed; the method specifically comprises the following steps:
and sequentially carrying out data format conversion, data segmentation processing and effective signal judgment on the read signal data to obtain an in-phase orthogonal signal.
As an improvement of the method, the input of the radio frequency fingerprint identification model is an in-phase orthogonal signal, the output is a radio frequency fingerprint identification result, a CNN-IQ network is adopted, the CNN-IQ network comprises a plurality of convolution layers and classification layers which are connected in sequence, wherein,
the multilayer convolutional layers comprise H convolutional layers, each convolutional layer expands an in-phase orthogonal signal into C channels through multi-channel filtering, and different characteristics are extracted in each channel; the output of each convolution layer is sequentially subjected to normalization processing and processing of an activation function, wherein the activation function is a ReLU; the 1 st convolutional layer adopts a (1,2) filter to extract IQ related characteristics, then adopts multilayer time domain filters to extract time domain characteristics, adopts maximum pooling to perform time domain direction data dimension reduction after every 2 time domain filters, then adopts self-adaptive average pooling, and inputs the characteristic mean value of C channels as a final characteristic value into a classification layer;
the classification layer adopts 1 full-connection layer, and the classification function is Sigmoid.
As an improvement of the above method, the training process of the radio frequency fingerprint identification model is as follows:
step 1), establishing a training set;
step 2) randomly selecting a plurality of data from a training set according to a certain proportion to respectively generate a training sample, a verification sample and a test sample;
step 3) carrying out supervised training on the training samples, carrying out 1-time verification every 2 times of training, and selecting the parameter which obtains the highest identification accuracy in the verification samples as the parameter of the radio frequency fingerprint identification model;
step 4) inputting the test sample into a radio frequency fingerprint identification model for testing to obtain model identification accuracy;
step 5) judging whether the model identification accuracy reaches a preset accuracy threshold, if not, adjusting parameters of the radio frequency fingerprint identification model, and turning to the step 3); and if so, obtaining the trained radio frequency fingerprint identification model.
As an improvement of the above method, the step 1) specifically includes:
marking serial numbers of the X Lora wireless transmitting modules as labels;
collecting communication signals of X Lora wireless transmitting modules;
preprocessing the acquired communication signals;
and associating the preprocessed communication signals with the serial number marks of the corresponding transmitting modules and establishing a training set.
A radio frequency fingerprint embedded real-time identification system based on a convolutional neural network, the system comprising: the system comprises a trained radio frequency fingerprint identification model, a signal receiving and collecting module, a real-time monitoring module, a preprocessing module and a radio frequency fingerprint identification module; wherein,
the signal receiving and collecting module is used for collecting wireless signals received by the antenna and storing files;
the real-time monitoring module is used for reading signal data from the file in sequence according to preset acquisition points and time length when a new wireless signal arrives through the change of the file;
the preprocessing module is used for preprocessing the read signal data;
and the radio frequency fingerprint identification module is used for inputting the preprocessed signal data into a pre-established radio frequency fingerprint identification model and outputting a radio frequency equipment label corresponding to the signal data to obtain a radio frequency fingerprint identification result.
As an improvement of the above system, the signal receiving and acquiring module comprises a receiving unit and a processing and storing unit; wherein;
the receiving unit is software defined radio equipment and is used for acquiring wireless signals received by the antenna, and outputting sampling signals to the processing and storing unit after processing;
the processing storage unit is realized based on a GNU Radio open source platform, and is used for performing low-pass filtering processing on the sampling signal and then performing file storage.
As an improvement of the above system, the specific implementation process of the preprocessing module is as follows:
and sequentially carrying out data format conversion, data segmentation processing and effective signal judgment on the read signal data to obtain an in-phase orthogonal signal.
Compared with the prior art, the invention has the advantages that:
1. the mobility and the real-time performance of the system provide guarantee for the wireless communication safety of mobile equipment (large-scale: satellites, airplanes, ships and the like; small-scale: mobile phones, unmanned aerial vehicles, robots and the like), and the wireless communication identity authentication of the mobile equipment is met by reducing the power consumption and the cost of the system;
2. the invention realizes the real-time radio frequency fingerprint identification of wireless signals by using the embedded development board with low cost and low power consumption, carrying SDR equipment, the light weight neural network and the real-time monitoring module.
Drawings
Fig. 1 is a technical flowchart of a radio frequency fingerprint embedded real-time identification method based on a convolutional neural network according to embodiment 1 of the present invention;
FIG. 2 is a flowchart of the RF fingerprinting process according to embodiment 1 of the invention;
fig. 3 is a schematic diagram of a CNN-IQ network structure according to embodiment 1 of the present invention;
FIG. 4 is a flowchart of the training of the radio frequency fingerprint identification model according to embodiment 1 of the present invention;
FIG. 5 is a block diagram of a system in which a signal receiving and collecting module employs a LimeSDR mini in embodiment 2 of the present invention;
FIG. 6 is a block diagram of a receiver with LimeSDR mini used as a signal receiving and collecting module in embodiment 2 of the present invention;
fig. 7 is a block diagram of a signal receiving and acquiring module using an LoRa wireless communication device according to embodiment 2 of the present invention;
fig. 8 is a GNU Radio signal acquisition GRC flow diagram of the signal receiving and acquiring module according to embodiment 2 of the present invention;
fig. 9 is a processing flow chart of the real-time monitoring module according to embodiment 2 of the present invention.
Detailed Description
The invention provides and realizes a neural network based on IQ two-dimensional characteristics of signals, greatly reduces the quantity of parameters and calculation amount, lightens the weight of the neural network, transplants the neural network into embedded equipment, reduces the cost of the whole system and enhances the mobility of an identification system. Meanwhile, the unmanned real-time monitoring function of the communication signals is completed, and real-time radio frequency fingerprint identification can be carried out.
The technical solution of the present invention will be described in detail below with reference to the accompanying drawings and examples.
Example 1
As shown in fig. 1, this is a flow chart of the whole rf fingerprint embedded real-time identification system. Firstly, a hardware platform of the system is an embedded development board of an arm architecture, the system is successfully verified by experiments on Jetson Nano and Raspberry type 4B in England, and a signal acquisition end is Software Defined Radio (SDR) equipment. And implanting the independently designed and trained convolutional neural network (CNN-IQ network) and a real-time monitoring functional module into a development board.
When a wireless signal is sent and received by the system, the real-time monitoring module detects the signal, the data reading and preprocessing module is called to preprocess the signal, the processed data are transmitted into a CNN-IQ network, and finally, a radio frequency fingerprint identification result is output by the network.
The specific treatment steps are as follows:
collecting and storing wireless signals received by an antenna;
when a new wireless signal is found, sequentially reading signal data from the storage according to a preset acquisition point number and a preset time length;
preprocessing the read signal data;
and inputting the preprocessed signal data into a pre-established radio frequency fingerprint identification model, and outputting a radio frequency equipment label corresponding to the signal data to obtain a radio frequency fingerprint identification result.
Radio frequency fingerprint identification model:
the invention provides an efficient convolutional neural network structure according to the characteristics of In-phase Quadrature signals (IQ for English abbreviation), and the efficient convolutional neural network structure is named as CNN-IQ. Fig. 2 is an end-to-end rf fingerprint identification process flow based on CNN-IQ with raw IQ sample data as input. The input of the network is an in-phase orthogonal signal, and the output is a radio frequency fingerprint identification result.
The CNN-IQ convolutional network structure is provided aiming at the problems of low feature utilization rate and large calculation amount when the current convolutional network processes sequential IQ signals. As shown in fig. 3, the data input is an IQ signal with a time length of N, the CNN-IQ first expands the input data into C channels (convolution kernels), and then performs extraction of IQ-related features and time-domain features in each channel, ensuring that each layer learns rich features; feature extraction extracts high-level, abstract features from H convolutional layers in addition to extracting features for each layer. And finally, only 1 full connection layer is adopted for classification. In the feature extraction layer, in each channel, the 1 st convolution layer adopts a (1,2) filter to extract IQ related features, the data dimension is changed from Nx 2 to Nx 1, and the calculation amount of subsequent processing is reduced by half; then extracting time domain characteristics by adopting a plurality of layers of small filters; meanwhile, in order to reduce the data length and achieve the purpose of reducing the calculated amount, the dimension of the time domain direction data is reduced by adopting maximum pooling after every 2 time domain filters; then, self-adaptive average pooling is adopted, and the characteristic mean value of each channel is used as a final characteristic value for classification; in the final classification and discrimination stage, only 1 full connection layer is adopted to reduce network parameters and calculation amount.
In addition, between the convolution output and the activation function, a Batch Normalization (BN) operation is added to increase the robustness and training speed of the model and replace dropout to prevent overfitting; reLU is adopted as an activation function in the network, and Sigmoid is adopted as a classification function in a classification layer.
The CNN-IQ network structure parameters are flexibly configurable. As shown in fig. 3, the CNN-IQ network is composed of H convolutional layers and 1 fully-connected layer, wherein each convolutional layer contains C feature extraction channels. To represent the CNN-IQ network with different structural parameters, the specific structural parameters are represented by CNN-IQ (H, C, S), where H represents the number of convolutional layers, C represents the number of convolutional cores (number of channels) per convolutional layer, and S represents the time domain convolutional core size. CNN-IQ network with 4 convolutional layers, each layer containing 32 time-domain convolution kernels of size (3,1), where the time-domain convolution kernel is of size (3,1), is represented as CNN-IQ (4,32,3). Table 1 is a network structure of CNN-IQ (4,32,3) with a data input format of 600 × 2.
TABLE 1 CNN-IQ (4,32,3) network architecture
| Layer name | Inputting size | Size/step | Number of convolution kernels |
| Conv2d-1 | 600×2 | 1×2/1×1 | 32 |
| BatchNorm2d-1 | 600×1 | - | 32 |
| MaxPool2d-1 | 600×1 | 2×1/2×1 | 32 |
| Conv2d-2 | 300×1 | 3×1/1×1 | 32 |
| BatchNorm2d-2 | 300×1 | - | 32 |
| Conv2d-3 | 300×1 | 3×1/1×1 | 32 |
| BatchNorm2d-3 | 300×1 | - | 32 |
| MaxPool2d-2 | 300×1 | 2×1/2×1 | 32 |
| Conv2d-4 | 150×1 | 3×1/1×1 | 32 |
| BatchNorm2d-4 | 150×1 | - | 32 |
| AdaptiveAvgPool2d | 150×1 | - | 32 |
| Linear-1 | 32 | - | - |
A preparation stage:
1. network training
The invention establishes a data set for signals needing Radio frequency fingerprint identification, collects the signals by using the communication signals of 113 Lora (English abbreviation of Long Range Radio, long distance Radio) wireless transmitting modules, and marks serial numbers 0-112 for each corresponding transmitting module. In order to increase the identification accuracy of the network, it is necessary to collect multiple sets of data in different environments as much as possible. The invention verifies that the data sets respectively collect 14 groups of signal data at indoor 1m and 10m and outdoor 10m and 60m, the single signal receiving time is 15s, and the total data set size is about 430G in total.
The single training, verifying and testing process of the neural network is shown in fig. 4, firstly, a non-return random sampling mode is adopted, a training sample, a verifying sample and a testing sample are generated by a data set according to the proportion of 7. The training round of the present invention was 30 times, except for the specific description.
2. Neural network and real-time module migration, hardware configuration
And transplanting the trained neural network parameters and the real-time module into an ARM development board. GNURadio is configured for different operating frequency bands for receiving SDR devices and wireless signals to be identified. And starts the real-time monitoring function.
The working stage is as follows:
when a wireless signal is sent and received by SDR equipment, GNURADio writes signal data into a system in real time, the real-time monitoring module detects the signal data, a file path of the signal data is output to a data reading and preprocessing module, the data reading and preprocessing module starts to read the written data, the signal is preprocessed after a preset data amount is read, the processed data is transmitted into a trained CNN-IQ network, and finally, a radio frequency fingerprint identification result is output by the network.
Example 2
the signal receiving and collecting module is used for collecting wireless signals received by the antenna and storing files;
the real-time monitoring module is used for reading signal data from the file in sequence according to the preset acquisition point number and the preset time length when the arrival of a new wireless signal is found through the change of the file;
the preprocessing module is used for preprocessing the read signal data;
and the radio frequency fingerprint identification module is used for inputting the preprocessed signal data into a pre-established radio frequency fingerprint identification model and outputting a radio frequency equipment label corresponding to the signal data to obtain a radio frequency fingerprint identification result.
The signal receiving and collecting module:
the signal receiving end is realized by a 3dB gain antenna and a LimeSDR-mini/RTL-SDR software defined radio, the radio frequency receiving principle is shown in figure 5, and the digital receiver framework is shown in figure 6. LimeSDR-mini uses LMS7002M as a receiving and transmitting chip of radio frequency signals, the LMS7002M is a highly programmable fully integrated, multi-band and multi-standard RF transceiver, signals received by a chip internal receiver block diagram shown in figure 6 directly enter a digital I/Q interface after passing through a transceiver signal processor, and are output to obtain I/Q signal sampling data.
A GNU Radio (open source software Radio development platform) open source platform and low-cost software defined Radio equipment (LimeSDR-mini/RTL-SDR) are used for building a signal acquisition system to acquire wireless signals. The structure of the acquisition system is shown in fig. 7, the acquisition system has flexible architecture, small volume and easy reconfiguration and redetection, and can meet the requirement of extracting the radio frequency fingerprint of a wireless communication signal. The GRC flow chart of GNU Radio software is shown in FIG. 8 (LimeSDR-mini is taken as an example).
A real-time monitoring module:
the purpose of the radio frequency fingerprint real-time identification system is to process received signals in real time and carry out reasoning and prediction. To achieve this, the recognition system needs to know when the signal arrives, then start reading the collected signal and pre-process, and finally send the data to the neural network for inference prediction.
As shown in fig. 9, the time of arrival of the signal is obtained by monitoring the creation time of the storage file of the signal data, because GNU Radio creates a new file for storage when acquiring new signal data. The monitoring of the creation of the storage file can enable the system to know when the signal arrives, and also enable the system to know the file path of data storage, and then the system starts to read data to a specified file and preprocesses the data, wherein the preprocessing mainly comprises data format conversion, data segmentation and effective signal judgment. The neural network model loads the trained parameters and performs inference prediction on the input data to identify a specific transmitter.
It should be noted that the present system is not limited to be applied to the embedded development board, but can also be applied to the PC motherboard.
The whole system is designed for the characteristic that signals of mobile equipment (large-scale: satellites, airplanes, ships and the like; small-scale: mobile phones, unmanned aerial vehicles, robots and the like) are highly sensitive, and wireless communication identity verification of the mobile equipment is met by reducing power consumption and cost of the system.
Finally, it should be noted that the above embodiments are only used for illustrating the technical solutions of the present invention and are not limited. Although the present invention has been described in detail with reference to the embodiments, it will be understood by those skilled in the art that various changes may be made and equivalents may be substituted without departing from the spirit and scope of the invention as defined in the appended claims.
Claims (6)
1. A radio frequency fingerprint embedded real-time identification method based on a convolutional neural network comprises the following steps:
collecting and storing wireless signals received by an antenna;
when a new wireless signal is found, reading signal data in sequence according to a preset acquisition point number and a preset time length;
preprocessing the read signal data;
inputting the preprocessed signal data into a pre-established radio frequency fingerprint identification model, and outputting a radio frequency equipment label corresponding to the signal data to obtain a radio frequency fingerprint identification result;
the wireless signal that collection antenna received carries out the storage, specifically includes:
collecting wireless signals received by an antenna, processing the wireless signals, outputting sampling signals, performing low-pass filtering processing, and then storing the sampling signals;
the input of the radio frequency fingerprint identification model is an in-phase orthogonal signal, the output is a radio frequency fingerprint identification result, a CNN-IQ network is adopted, the CNN-IQ network comprises a plurality of convolution layers and a classification layer which are connected in sequence, wherein,
the multilayer convolutional layers comprise H convolutional layers, each convolutional layer expands an in-phase orthogonal signal into C channels through multi-channel filtering, and different characteristics are extracted in each channel; the output of each convolution layer is sequentially subjected to normalization processing and processing of an activation function, wherein the activation function is a ReLU; the 1 st convolutional layer adopts a (1,2) filter to extract IQ related characteristics, then adopts multilayer time domain filters to extract time domain characteristics, adopts maximum pooling to perform time domain direction data dimension reduction after every 2 time domain filters, then adopts self-adaptive average pooling, and inputs the characteristic mean value of C channels as a final characteristic value into a classification layer;
the classification layer adopts 1 full-connection layer, and the classification function is Sigmoid.
2. The embedded real-time identification method for radio frequency fingerprints based on convolutional neural network as claimed in claim 1, wherein the read signal data is preprocessed; the method specifically comprises the following steps:
and sequentially carrying out data format conversion, data segmentation processing and effective signal judgment on the read signal data to obtain an in-phase orthogonal signal.
3. The embedded real-time radio frequency fingerprint identification method based on the convolutional neural network as claimed in claim 1, wherein the training process of the radio frequency fingerprint identification model is as follows:
step 1), establishing a training set;
step 2) randomly selecting a plurality of data from a training set according to a certain proportion to respectively generate a training sample, a verification sample and a test sample;
step 3) performing supervision training on the training samples, performing verification for 1 time in each training 2 rounds, and selecting the parameter which obtains the highest identification accuracy in the verification samples as the parameter of the radio frequency fingerprint identification model;
step 4) inputting the test sample into a radio frequency fingerprint identification model for testing to obtain model identification accuracy;
step 5) judging whether the model identification accuracy reaches a preset accuracy threshold, if not, adjusting parameters of the radio frequency fingerprint identification model, and turning to the step 3); and if so, obtaining the trained radio frequency fingerprint identification model.
4. The embedded real-time radio frequency fingerprint identification method based on the convolutional neural network as claimed in claim 3, wherein the step 1) specifically comprises:
marking serial numbers of the X Lora wireless transmitting modules as labels;
collecting communication signals of X Lora wireless transmitting modules;
preprocessing the acquired communication signals;
and associating the preprocessed communication signals with the serial number marks of the corresponding transmitting modules and establishing a training set.
5. A radio frequency fingerprint embedded real-time identification system based on a convolutional neural network is characterized by comprising: the system comprises a trained radio frequency fingerprint identification model, a signal receiving and collecting module, a real-time monitoring module, a preprocessing module and a radio frequency fingerprint identification module; wherein,
the signal receiving and collecting module is used for collecting wireless signals received by the antenna and storing files;
the real-time monitoring module is used for reading signal data from the file in sequence according to preset acquisition points and time length when a new wireless signal arrives through the change of the file;
the preprocessing module is used for preprocessing the read signal data;
the radio frequency fingerprint identification module is used for inputting the preprocessed signal data into a pre-established radio frequency fingerprint identification model and outputting a radio frequency equipment label corresponding to the signal data to obtain a radio frequency fingerprint identification result;
the signal receiving and acquiring module comprises a receiving unit and a processing and storing unit; wherein;
the receiving unit is software defined radio equipment and is used for acquiring wireless signals received by the antenna, and outputting sampling signals to the processing and storing unit after processing;
the processing storage unit is realized based on a GNU Radio open source platform, and is used for performing low-pass filtering processing on the sampling signal and then performing file storage;
the input of the radio frequency fingerprint identification model is an in-phase orthogonal signal, the output is a radio frequency fingerprint identification result, a CNN-IQ network is adopted, the CNN-IQ network comprises a plurality of convolution layers and a classification layer which are connected in sequence, wherein,
the multilayer convolutional layers comprise H convolutional layers, each convolutional layer expands an in-phase orthogonal signal into C channels through multi-channel filtering, and different characteristics are extracted in each channel; the output of each convolution layer is sequentially subjected to normalization processing and processing of an activation function, wherein the activation function is ReLU; the 1 st convolutional layer adopts a (1,2) filter to extract IQ related characteristics, then adopts multilayer time domain filters to extract time domain characteristics, adopts maximum pooling to perform time domain direction data dimension reduction after every 2 time domain filters, then adopts self-adaptive average pooling, and inputs the characteristic mean value of C channels as a final characteristic value into a classification layer;
the classification layer adopts 1 full-connection layer, and the classification function is Sigmoid.
6. The radio frequency fingerprint embedded real-time identification system based on the convolutional neural network of claim 5, wherein the pre-processing module is specifically implemented by the following processes:
and sequentially carrying out data format conversion, data segmentation processing and effective signal judgment on the read signal data to obtain an in-phase orthogonal signal.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110048449.5A CN112867010B (en) | 2021-01-14 | 2021-01-14 | Radio frequency fingerprint embedded real-time identification method and system based on convolutional neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110048449.5A CN112867010B (en) | 2021-01-14 | 2021-01-14 | Radio frequency fingerprint embedded real-time identification method and system based on convolutional neural network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112867010A CN112867010A (en) | 2021-05-28 |
| CN112867010B true CN112867010B (en) | 2023-04-18 |
Family
ID=76005762
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110048449.5A Active CN112867010B (en) | 2021-01-14 | 2021-01-14 | Radio frequency fingerprint embedded real-time identification method and system based on convolutional neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112867010B (en) |
Families Citing this family (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113949986B (en) * | 2021-08-11 | 2024-03-29 | 厦门大学 | Positioning method, positioning system and monitoring method based on WR (Wireless Fidelity) and RF (radio frequency) fingerprint technology |
| CN114387627A (en) * | 2022-01-11 | 2022-04-22 | 厦门大学 | Small sample wireless device radio frequency fingerprint identification method and device based on depth measurement learning |
| CN114997299B (en) * | 2022-05-27 | 2024-04-16 | 电子科技大学 | Radio frequency fingerprint identification method in resource limited environment |
| CN116776227B (en) * | 2023-08-04 | 2024-07-30 | 中国人民解放军战略支援部队航天工程大学 | Satellite identity recognition method and device based on feature fusion |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110009051A (en) * | 2019-04-11 | 2019-07-12 | 浙江立元通信技术股份有限公司 | Feature extraction unit and method, DCNN model, recognition methods and medium |
| CN110855591A (en) * | 2019-12-09 | 2020-02-28 | 山东大学 | QAM and PSK signal intra-class modulation classification method based on convolutional neural network structure |
| CN111368850A (en) * | 2018-12-25 | 2020-07-03 | 展讯通信(天津)有限公司 | Image feature extraction method, image target detection method, image feature extraction device, image target detection device, convolution device, CNN network device and terminal |
| CN111382833A (en) * | 2018-12-29 | 2020-07-07 | 佳能株式会社 | Method and device for training and applying multilayer neural network model and storage medium |
| CN111639677A (en) * | 2020-05-07 | 2020-09-08 | 齐齐哈尔大学 | Garbage image classification method based on multi-branch channel capacity expansion network |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US11640522B2 (en) * | 2018-12-13 | 2023-05-02 | Tybalt, Llc | Computational efficiency improvements for artificial neural networks |
-
2021
- 2021-01-14 CN CN202110048449.5A patent/CN112867010B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111368850A (en) * | 2018-12-25 | 2020-07-03 | 展讯通信(天津)有限公司 | Image feature extraction method, image target detection method, image feature extraction device, image target detection device, convolution device, CNN network device and terminal |
| CN111382833A (en) * | 2018-12-29 | 2020-07-07 | 佳能株式会社 | Method and device for training and applying multilayer neural network model and storage medium |
| CN110009051A (en) * | 2019-04-11 | 2019-07-12 | 浙江立元通信技术股份有限公司 | Feature extraction unit and method, DCNN model, recognition methods and medium |
| CN110855591A (en) * | 2019-12-09 | 2020-02-28 | 山东大学 | QAM and PSK signal intra-class modulation classification method based on convolutional neural network structure |
| CN111639677A (en) * | 2020-05-07 | 2020-09-08 | 齐齐哈尔大学 | Garbage image classification method based on multi-branch channel capacity expansion network |
Non-Patent Citations (3)
| Title |
|---|
| Deep Learning Convolutional Neural Networks for Radio Identification;Shamnaz Riyaz 等;《IEEE Communications Magazine 》;20180930;第56卷(第9期);第146、149-150页 * |
| Emitter Identification Using CNN IQ Imbalance Estimators;Lauren J. Wong等;《ARXIV》;全文 * |
| Radio Frequency Fingerprint Identification for LoRa Using Spectrogram and CNN;Guanxiong Shen等;《ARXIV》;全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112867010A (en) | 2021-05-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112867010B (en) | Radio frequency fingerprint embedded real-time identification method and system based on convolutional neural network | |
| CN112702294B (en) | Modulation recognition method for multi-level feature extraction based on deep learning | |
| CN108509911B (en) | Interference signal identification method based on convolutional neural network | |
| CN108960138B (en) | Equipment authentication feature identification method based on convolutional neural network | |
| US11700156B1 (en) | Intelligent data and knowledge-driven method for modulation recognition | |
| CN111310700A (en) | Intermediate frequency sampling sequence processing method for radiation source fingerprint feature identification | |
| CN112183279B (en) | Communication radiation source individual identification method based on IQ graph characteristics | |
| CN116257750A (en) | Radio frequency fingerprint identification method based on sample enhancement and deep learning | |
| CN114417914A (en) | Radio frequency fingerprint extraction and identification method based on multi-channel convolutional neural network | |
| CN114980122A (en) | Small sample radio frequency fingerprint intelligent identification system and method | |
| CN110087233B (en) | Unmanned aerial vehicle identity authentication method based on radio frequency fingerprint | |
| CN113807166B (en) | Image processing method, device and storage medium | |
| CN111027614A (en) | Noise-enhanced radio frequency fingerprint identification method and device | |
| CN116776227B (en) | Satellite identity recognition method and device based on feature fusion | |
| CN113347637B (en) | Radio Frequency (RF) fingerprint identification method and device based on embedded wireless equipment | |
| Feng et al. | FCGCN: Feature Correlation Graph Convolution Network for Few-Shot Individual Identification | |
| Yuan et al. | Wireless device identification based on improved convolutional neural network model | |
| CN116580290B (en) | Unmanned aerial vehicle identification method, unmanned aerial vehicle identification device and storage medium | |
| CN115809426A (en) | Radiation source individual identification method and system | |
| CN112801910B (en) | Channel state information image denoising method and indoor positioning model | |
| Qu et al. | Specific emitter identification based on one-dimensional complex-valued residual networks with an attention mechanism | |
| Gu et al. | Radio frequency fingerprinting driven drone identification based on complex-valued CNN | |
| CN113705646A (en) | Radio frequency fine characteristic information extraction method and system based on semi-supervised meta learning | |
| CN111917674A (en) | Modulation identification method based on deep learning | |
| CN112232430A (en) | Neural network model testing method and device, storage medium and electronic device |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |