CN114417914A - Radio frequency fingerprint extraction and identification method based on multi-channel convolutional neural network - Google Patents
Radio frequency fingerprint extraction and identification method based on multi-channel convolutional neural network Download PDFInfo
- Publication number
- CN114417914A CN114417914A CN202111624298.XA CN202111624298A CN114417914A CN 114417914 A CN114417914 A CN 114417914A CN 202111624298 A CN202111624298 A CN 202111624298A CN 114417914 A CN114417914 A CN 114417914A
- Authority
- CN
- China
- Prior art keywords
- signal
- neural network
- convolutional neural
- channel
- radio frequency
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/12—Classification; Matching
-
- 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
- G06N3/045—Combinations of networks
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W12/00—Security arrangements; Authentication; Protecting privacy or anonymity
- H04W12/60—Context-dependent security
- H04W12/69—Identity-dependent
- H04W12/79—Radio fingerprint
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Computer Security & Cryptography (AREA)
- Signal Processing (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Mathematical Physics (AREA)
- Data Mining & Analysis (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Evolutionary Computation (AREA)
- Software Systems (AREA)
- Biomedical Technology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Computer Networks & Wireless Communication (AREA)
- Mobile Radio Communication Systems (AREA)
Abstract
The invention relates to the technical field of information security, in particular to a radio frequency fingerprint extraction and identification method based on a multi-channel convolutional neural network. The method comprises the following steps: (1) collecting a sending signal of wireless equipment and preprocessing the sending signal; (2) for each signal, dividing the signal into a front transient part, a steady state part and a tail transient part, and respectively generating a differential constellation trajectory Diagram (DCTF); (3) dividing the obtained DCTF data set into a training set and a test set, and training a multi-channel convolutional neural network classifier by using the training set; (4) and for the received signals sent by the wireless equipment, carrying out classification and identification on the signals by using a trained multi-channel convolutional neural network. The invention constructs a multi-channel convolution neural network by using DCTF of different parts of signals to extract and identify the radio frequency fingerprint, avoids partial signal characteristics from being submerged in the DCTF and balances the characteristic weights of different parts of signals, and ensures that the identification effect is better, especially in the scene of low signal-to-noise ratio.
Description
Technical Field
The invention relates to the technical field of information security, in particular to a radio frequency fingerprint extraction and identification method based on a multi-channel convolutional neural network.
Background
Wireless communication technology has become an integral part of modern life. However, the broadcast nature of wireless communication transmissions makes mobile devices more vulnerable to malicious attacks. Radio frequency fingerprinting is a physical layer solution for mobile device identification and authentication. Due to the limitations of the manufacturing process, electronic components may have small defects and imperfections, and these small deviations may result in distortions or distortions of some signals that may be used to detect the equipment transmitting the signals. By extracting and identifying the radio frequency fingerprint of the mobile equipment, the safety of the wireless communication system is enhanced.
The most similar solutions to the present invention are found in the literature: S.Wang, L.Peng, H.Fu, A.Hu and X.Zhou, "A capacitive Neural Network-Based RF converting Identification Scheme for Mobile Phones," IEEE INFOCOM 2020-IEEE Conference on Computer Communications workstations (INFOCOM WKSHPS),2020, pp.115-120, doi: 10.1109/INFOCOMWKSHPS50562.2020.9163058, the main idea is to obtain its statistical properties by transforming the whole segment of the received signal into a differential constellation diagram, then train a Convolutional Neural Network with it, and finally input the received signal into the Network for Identification.
In the above scheme, in the process of converting the whole signal into the differential constellation locus diagram, the covering relationship of the transient and steady-state part pixels of the signal in the differential constellation locus diagram is ignored, and the contribution of the transient part due to the difference of the number of sampling points of the transient and steady-state parts of the signal is reduced in the differential constellation locus diagram, thereby affecting the classification and identification effects.
List of abbreviations, english and key term definitions:
1. DCTF differential constellation Trace figure
2. I/Q signals: in-phase/quadrature signal
3. USRP: universal Software Radio Peripheral
Disclosure of Invention
In order to solve the technical problems in the background art, the invention aims to provide a radio frequency fingerprint extraction and identification method based on a multi-channel convolutional neural network.
In order to achieve the technical purpose, the technical scheme of the invention is as follows:
the radio frequency fingerprint extraction and identification method based on the multi-channel convolutional neural network comprises the following steps:
(1) collecting a sending signal of wireless equipment and preprocessing the sending signal;
(2) for each signal, dividing the signal into a front transient part, a steady state part and a tail transient part, and respectively generating a differential constellation locus diagram;
(3) dividing the obtained differential constellation trajectory diagram data set into a training set and a testing set, and training a multi-channel convolutional neural network classifier by using the training set;
(4) and for the received signals sent by the wireless equipment, carrying out classification and identification on the signals by using a trained multi-channel convolutional neural network.
Further, when the receiving end collects the baseband signals of the wireless equipment in the step (1), the sampling rate meets the nyquist law.
Further, the pretreatment in step (1) comprises: signal detection and synchronization, energy normalization.
Further, in the step (2), the front transient is a signal part of the device starting to send a signal reaching the rated power, the tail transient is a signal part of the device starting to drop in power until the device stops working, and the steady-state part is a signal part except the signal part.
Further, in the step (2), difference processing is respectively carried out on the obtained front transient state, steady state and tail transient state parts of the signals according to the selected sampling point intervals, the I/Q signals after difference processing are drawn on a two-dimensional plane with the I path and the Q path as coordinate axes, and the gray value is mapped according to the density of the sampling points to obtain a difference constellation locus diagram.
Further, in step (3), the multichannel convolutional neural network structure is composed of the following parts:
differential constellation trajectory diagram input stage: and (3) inputting the differential constellation locus diagrams reflecting different part characteristics of the signal in the step (2) into different channels of the convolutional neural network, wherein each channel corresponds to one differential constellation locus diagram.
A characteristic extraction stage: the difference constellation locus diagram of each channel is input into a convolution branch with the same structure as LeNet-5, and each convolution branch consists of two convolution layers and two pooling layers.
A classification stage: all matrices in different channels are flattened into vectors and connected into one eigenvector. And then inputting the vector into full connection layers to integrate local feature information obtained from each channel for classification, wherein the number of the full connection layers is three, and the probability of each label is obtained by using a softmax function during classification.
Further, in step (4), the received wireless device signal is processed according to steps (1) and (2) to obtain a differential constellation locus diagram of a signal front transient state, a signal steady state and a signal tail transient state, the differential constellation locus diagram is input to the multi-channel convolutional neural network obtained in step (3), a probability value of each label is output, and the label with the maximum probability value is a prediction result.
Adopt the beneficial effect that above-mentioned technical scheme brought:
compared with the prior art, the invention has the following remarkable advantages: the characteristic loss caused by mutual covering of transient part pixels and steady part pixels when the signals are converted into the differential constellation locus diagram is avoided, and the weight of the influence of different part characteristics of the signals on the radio frequency fingerprint is balanced. Experiments show that compared with the conventional convolution neural network method based on the differential constellation locus diagram, the method disclosed by the invention can further improve the identification performance of the wireless equipment, especially in a low signal-to-noise ratio scene.
Drawings
FIG. 1 is a schematic flow diagram of one embodiment of the present invention;
FIG. 2 is an example of an experimental platform according to an embodiment of the present invention;
fig. 3 is a DCTF diagram of random access signals of different LTE terminals;
fig. 4 is a DCTF diagram of a pre-transient, steady-state, and tail transient of an LTE terminal random access signal;
FIG. 5 is a diagram of a multi-channel convolutional neural network architecture;
FIG. 6 is a schematic diagram of performance improvement of radio frequency fingerprint classification and identification based on a multi-channel convolutional neural network.
Detailed Description
The embodiment provides a radio frequency fingerprint extraction and identification method based on a multi-channel convolutional neural network, as shown in fig. 1, the method comprises the following steps:
(1) and collecting a sending signal of the wireless equipment and preprocessing the sending signal.
In this embodiment, as shown in fig. 2, an experimental platform is built and signals are collected, specifically:
and 6 LTE mobile phone devices are selected as target wireless devices and are numbered. An LTE pseudo base station is built by utilizing open source platform software and one USRPB205 device, the uplink central frequency point of the pseudo base station is 2565MHz, the number of resource blocks is 50, and the central frequency point of a random access signal set by the pseudo base station is 2561.4 MHz. The mobile phone is connected with the pseudo base station, and the flying mode of the mobile phone is triggered and closed continuously and manually to acquire the random access signal of the mobile phone.
The signal is collected by using one USRPB205 device, the central frequency point of the sampling is 2561.4MHz, the sampling rate is 16M, and the number of the collected sampling points of the random access signal of each LTE mobile phone is 14450. The sampling environment is the sight distance and the signal-to-noise ratio is high. In order to explore the performance of the invention under different signal-to-noise ratios, Gaussian white noise is added to the collected signals to make the signal-to-noise ratio of the collected signals to be from 10dB to 30dB, and each collected signal is simulated once according to the steps (2) to (4) at each signal-to-noise ratio, wherein each group is divided by 5 dB.
The preprocessing comprises signal detection and synchronization and energy normalization. Detecting the existence of signal by using signal spectrum energy, carrying out coarse synchronization on received signal by using cyclic prefix, and then using local standardThe random access signal fine synchronizes the received signal. Then, the signal is subjected to energy normalization processing to obtain a signalWherein alpha iscFor the channel attenuation factor, s (t) is the baseband signal, Δ f is the deviation of the sampling frequency from the signal transmission frequency, φ is the phase deviation, and A is the root mean square of the received signal.
(2) For each signal, the signal is divided into a front transient part, a steady state part and a tail transient part, and a differential constellation locus diagram is generated respectively.
The dividing of the signal into the front transient state, the steady state and the tail transient state is specifically to divide 50 sampling points before the signal is sent and 50 sampling points after the signal is sent into the front transient state, divide 50 sampling points before the signal is ended and 50 sampling points after the signal is ended into the tail transient state according to the signal energy characteristics of the LTE random access signal, and divide the middle part of the signal except the above parts into the steady state. And respectively generating a differential constellation locus diagram for the three parts. Fig. 3 shows a differential constellation trace diagram of a complete LTE random access signal, and fig. 4 shows a differential constellation trace diagram of a front transient state, a steady state, and a tail transient state of an LTE random access signal.
The method for generating the differential constellation locus diagram is as follows:
carrying out difference operation on the signal Y (t) obtained in the step (1):
wherein (g)*For conjugate operation, tnIs a differential interval. After the difference processing, a stable phase rotation factor is obtainedTherefore, a stable differential constellation locus diagram can be obtained.
Drawing the I/Q signals after the difference processing on a two-dimensional plane with the I path and the Q path as coordinate axes, and mapping a gray value according to the density of sampling points to obtain a differential constellation locus diagram.
(3) And dividing the obtained differential constellation trajectory diagram data set into a training set and a testing set, and training a multi-channel convolutional neural network classifier by using the training set.
And (3) the difference constellation trajectory diagram data set is processed according to the following steps of 5: 1, randomly dividing the ratio into a training set and a testing set, and adopting a hierarchical sampling method to ensure that the ratio of each label in the training set and the testing set is the same as that in the data set. And training a multi-channel convolutional neural network by using the training set.
The structure of the multichannel convolutional neural network is shown in figure 5. The structure of the device is composed of the following parts:
differential constellation trajectory diagram input stage: and (3) inputting the differential constellation locus diagrams reflecting different part characteristics of the signal in the step (2) into different channels of the convolutional neural network, wherein each channel corresponds to one differential constellation locus diagram. The size of each differential constellation trace is 129 x 129.
A characteristic extraction stage: the difference constellation locus diagram of each channel is input into a convolution branch with the same structure as LeNet-5, and each convolution branch consists of two convolution layers and two pooling layers. The convolution layer is used to extract the local features of the DCTF, and the size of the convolution kernel is 5 x 5. Pooling layers are added after the convolutional layer to reduce the number of parameters to speed up the computation and prevent overfitting. In this example, maximum pooling is used with a step size of 2 and a pooling size of 2 x 2.
A classification stage: all matrices in different channels are flattened into vectors and connected into one eigenvector. And then inputting the vector into full connection layers to integrate local feature information obtained from each channel for classification, wherein the number of the full connection layers is three, and the probability of each label is obtained by using a softmax function during classification.
(4) And for the received signals sent by the wireless equipment, carrying out classification and identification on the signals by using a trained multi-channel convolutional neural network.
And (3) processing the received wireless equipment signals according to the step (1) and the step (2) to obtain a differential constellation locus diagram of a signal front transient state, a steady state and a tail transient state, inputting the differential constellation locus diagram into the multi-channel convolutional neural network obtained in the step (3), and outputting the probability value of each label, wherein the label with the maximum probability value is a prediction result.
By the method, the characteristic loss caused by mutual covering of transient part pixels and steady part pixels when the signals are converted into the differential constellation locus diagram is avoided, the weight of influence of different part characteristics of the signals on the radio frequency fingerprint is balanced, and the accuracy of wireless equipment identification can be effectively improved. FIG. 6 is a test result of classifying a test set using and without the present method for direct convolutional neural networks. It can be seen that when the method of the present invention is used, the recognition accuracy is improved under all the tested signal-to-noise ratios (10dB-30 dB); especially under the scene of low signal-to-noise ratio (less than or equal to 15dB), the improvement effect is more obvious.
The above disclosure is only one preferred embodiment of the present invention, and the scope of the present invention should not be limited thereby, and the present invention is covered by the claims.
Those not described in detail in this specification are within the skill of the art.
Claims (7)
1. A radio frequency fingerprint extraction and identification method based on a multi-channel convolution neural network is characterized by comprising the following steps:
(1) collecting a sending signal of wireless equipment and preprocessing the sending signal;
(2) for each signal, dividing the signal into a front transient part, a steady state part and a tail transient part, and respectively generating a differential constellation locus diagram;
(3) dividing the obtained differential constellation trajectory diagram data set into a training set and a testing set, and training a multi-channel convolutional neural network classifier by using the training set;
(4) and for the received signals sent by the wireless equipment, carrying out classification and identification on the signals by using a trained multi-channel convolutional neural network.
2. The method for extracting and identifying radio frequency fingerprints based on the multi-channel convolutional neural network as claimed in claim 1, wherein the sampling rate satisfies nyquist's law when the receiving end collects baseband signals of the wireless device in step (1).
3. The method for extracting and identifying the radio frequency fingerprint based on the multi-channel convolutional neural network as claimed in claim 1, wherein the preprocessing in step (1) comprises: signal detection and synchronization, energy normalization.
4. The method for extracting and identifying radio frequency fingerprint based on multi-channel convolutional neural network of claim 1, wherein the front transient in step (2) is a signal portion where the device starts to send signal to reach the rated power, the tail transient is a signal portion where the device power starts to drop until the device stops working, and the steady state portion is a signal portion except the above portion.
5. The method for extracting and identifying radio frequency fingerprints based on the multichannel convolutional neural network as claimed in claim 1, wherein in the step (2), the obtained front transient part, the steady state part and the tail transient part of the signal are respectively subjected to differential processing according to selected sampling point intervals, the I/Q signal after the differential processing is drawn on a two-dimensional plane with an I path and a Q path as coordinate axes, and a gray value is mapped according to the density of the sampling points to obtain a differential constellation locus diagram.
6. The method for extracting and identifying radio frequency fingerprint based on multi-channel convolutional neural network as claimed in claim 1, wherein in step (3), the multi-channel convolutional neural network structure is composed of the following parts:
differential constellation trajectory diagram input stage: inputting the differential constellation locus diagrams reflecting different part characteristics of the signals in the step (2) into different channels of the convolutional neural network, wherein each channel corresponds to one differential constellation locus diagram;
a characteristic extraction stage: inputting the differential constellation locus diagram of each channel into a convolution branch with the same structure as LeNet-5, wherein each convolution branch consists of two convolution layers and two pooling layers;
a classification stage: flattening all the matrixes in different channels into vectors, connecting the vectors into a feature vector, inputting the vectors into a full connection layer to integrate local feature information obtained from each channel for classification, wherein the number of the full connection layers is three, and the probability of each label is obtained by using a softmax function during classification.
7. The method for extracting and identifying radio frequency fingerprints based on the multi-channel convolutional neural network as claimed in claim 1, wherein in step (4), the received wireless device signals are processed according to steps (1) and (2) to obtain a differential constellation locus diagram of a signal front transient state, a signal steady state and a signal tail transient state, the differential constellation locus diagram is input into the multi-channel convolutional neural network obtained in step (3), a probability value of each label is output, the label with the maximum probability value is a prediction result, and the probability value is the maximum label.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111624298.XA CN114417914A (en) | 2021-12-28 | 2021-12-28 | Radio frequency fingerprint extraction and identification method based on multi-channel convolutional neural network |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111624298.XA CN114417914A (en) | 2021-12-28 | 2021-12-28 | Radio frequency fingerprint extraction and identification method based on multi-channel convolutional neural network |
Publications (1)
Publication Number | Publication Date |
---|---|
CN114417914A true CN114417914A (en) | 2022-04-29 |
Family
ID=81270417
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202111624298.XA Pending CN114417914A (en) | 2021-12-28 | 2021-12-28 | Radio frequency fingerprint extraction and identification method based on multi-channel convolutional neural network |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN114417914A (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN115314348A (en) * | 2022-08-03 | 2022-11-08 | 电信科学技术第五研究所有限公司 | Convolutional neural network-based QAM signal modulation identification method |
CN116127298A (en) * | 2023-02-22 | 2023-05-16 | 北京邮电大学 | Small sample radio frequency fingerprint identification method based on triplet loss |
-
2021
- 2021-12-28 CN CN202111624298.XA patent/CN114417914A/en active Pending
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN115314348A (en) * | 2022-08-03 | 2022-11-08 | 电信科学技术第五研究所有限公司 | Convolutional neural network-based QAM signal modulation identification method |
CN115314348B (en) * | 2022-08-03 | 2023-10-24 | 电信科学技术第五研究所有限公司 | QAM signal modulation identification method based on convolutional neural network |
CN116127298A (en) * | 2023-02-22 | 2023-05-16 | 北京邮电大学 | Small sample radio frequency fingerprint identification method based on triplet loss |
CN116127298B (en) * | 2023-02-22 | 2024-03-19 | 北京邮电大学 | Small sample radio frequency fingerprint identification method based on triplet loss |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Yang et al. | Specific emitter identification with limited samples: A model-agnostic meta-learning approach | |
CN114417914A (en) | Radio frequency fingerprint extraction and identification method based on multi-channel convolutional neural network | |
CN108960138B (en) | Equipment authentication feature identification method based on convolutional neural network | |
CN108764077B (en) | Digital signal modulation classification method based on convolutional neural network | |
CN106845339B (en) | Mobile phone individual identification method based on bispectrum and EMD fusion characteristics | |
CN107979842B (en) | A kind of mobile phone individual discrimination method based on bispectrum feature and deep learning | |
Hanna et al. | Deep learning based transmitter identification using power amplifier nonlinearity | |
CN109921886B (en) | Robust low-power-consumption equipment radio frequency fingerprint identification method | |
CN112867010B (en) | Radio frequency fingerprint embedded real-time identification method and system based on convolutional neural network | |
CN110351766B (en) | Method for carrying out multi-cell blind detection and measurement processing aiming at 5G NR | |
CN111601307A (en) | Transient-steady state based partial superposition radio frequency fingerprint method | |
Ren et al. | A practical study of channel influence on radio frequency fingerprint features | |
CN113225282A (en) | Communication signal modulation identification method based on BP neural network | |
Zhang et al. | Deep radio fingerprint resnet for reliable lightweight device identification | |
Yang et al. | Deep learning based RFF recognition with differential constellation trace figure towards closed and open set | |
CN115296759A (en) | Interference identification method based on deep learning | |
CN111027614A (en) | Noise-enhanced radio frequency fingerprint identification method and device | |
Qiu et al. | Signal-independent RFF identification for LTE mobile devices via ensemble deep learning | |
CN113098638B (en) | Weak signal detection method based on grouped range diagram | |
Dong et al. | Radar signal modulation recognition based on bispectrum features and deep learning | |
Wang et al. | Specific emitter identification based on deep adversarial domain adaptation | |
CN113343868A (en) | Radiation source individual identification method and device, terminal and storage medium | |
CN113905383B (en) | IFF signal identification method, device and medium based on radio frequency fingerprint | |
CN115809426A (en) | Radiation source individual identification method and system | |
Makled et al. | Identification of cellular signal measurements using machine learning |
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 |