Wi-Fi radio frequency fingerprint identification method and device based on domain self-adaption

Abstract

The invention discloses a Wi-Fi (Wireless Fidelity) radio frequency fingerprint identification method and device based on domain self-adaption. The method comprises the following steps of: 1) acquiring and preprocessing data; 2) data enhancement: based on the signal, providing a data random enhancement method based on physical damage to carry out random frequency and phase enhancement on training data; 3) feature extraction: designing a method of combining steady state and transient state signal logarithmic spectrums based on the enhanced data, and extracting radio frequency fingerprint features as model input; 4) constructing a domain adversarial transfer learning network model; 5) training the model and evaluating a detection result; and 6) model application: performing identification application on the trained model in an actual electromagnetic scene. According to the method provided by the invention, an adversarial transfer learning network mechanism is introduced into the radio frequency fingerprint field, and meanwhile, a physical damage-based random data enhancement method and a channel robust time-frequency logarithmic spectrum feature are designed, so that the recognition precision of an existing model on unknown target domain data is successfully improved; experiments show that the accuracy of the method is improved by about 10%-25% compared with that of an existing model in a single-domain generalization scene, and the problem of data distribution drift of the existing model across different temperatures and channel states is solved.

Description

Technical Field

[0001] The present invention relates to the field of signal security technology, and in particular to a Wi-Fi radio frequency fingerprint recognition method and device based on domain adaptation. Background Art

[0002] In recent years, with the rapid development of industrial Internet of Things, wireless networks and radio frequency signals have been widely used in industrial manufacturing, smart cities, commerce and military fields. However, most of the existing wireless network authentication methods are based on link layer protocols, which are vulnerable to forgery and tampering attacks. For example, attackers can disguise themselves as normal users to access the network by modifying device identifiers such as network IP or MAC addresses, thereby stealing network data, leading to disastrous consequences. Therefore, a more stable and effective network security authentication method is needed.

[0003] Although the physical layer radio frequency fingerprint (RF) can effectively deal with this network attack, due to the complex electromagnetic environment in real scenes and the influence of temperature, channel status, etc., the robustness of RF fingerprint features is poor and it is difficult to cope with different cross-domain scenarios. In response to this problem, domestic and foreign scholars have carried out a variety of research, which can be mainly divided into: 1) Fingerprint extraction methods based on traditional transient and steady-state signals: mainly including transient signal characteristics (amplitude envelope, phase, etc.) and steady-state characteristics (carrier frequency deviation, steady-state spectrum, etc.), and extracting transient fingerprints through a series of feature transformations for device identification; 2) Fingerprint extraction methods based on deep learning: mainly extracting implicit signal features through neural networks, and then realizing the identification and classification of RF fingerprints through classification networks, and improving the robustness to different cross-domain scenarios through network parameter optimization. However, these RF fingerprint recognition methods all have certain defects: for example, the RF fingerprints extracted based on traditional transient and steady-state signal methods rely more on expert knowledge and tend to ignore potential implicit features in the signal, resulting in low recognition accuracy; and although the existing fingerprint extraction methods based on deep learning can mine implicit features in the signal, in complex real electromagnetic environments, the distribution of original data will shift, resulting in poor model robustness and difficulty in coping with complex cross-domain scenarios.

[0004] Therefore, the existing technology has not achieved obvious generalization effect in radio frequency fingerprint recognition. In order to solve these problems, it is urgent to provide a robust radio frequency fingerprint recognition method in the current complex real electromagnetic environment. Summary of the invention

[0005] In order to overcome the shortcomings of existing identification methods, the present invention proposes a Wi-Fi radio frequency fingerprint identification method and device based on domain adaptation, which is used to solve the technical problem that the recognition accuracy of the deep learning model in the existing identification method decreases significantly with the change of temperature or channel in the environment.

[0006] The method proposed in the present invention successfully improves the recognition accuracy of the existing model for unknown target domain data by introducing an adversarial transfer learning network mechanism in the field of RF fingerprinting, and designs a random data enhancement method based on physical damage and a channel-robust time-frequency logarithmic spectrum feature. Experiments show that the method improves the accuracy of the existing model by about 10%-25% in a single-domain generalization scenario, and solves the problem of data distribution drift of the existing model across different temperatures and channel states.

[0007] The technical solution adopted by the present invention to solve its technical problem is:

[0008] A Wi-Fi radio frequency fingerprint recognition method based on domain adaptation, the method comprising the following steps:

[0009] Step S1: Data collection and preprocessing: Construct a Wi-Fi signal detector to collect radio signals from the air, detect the Wi-Fi signal from it through time synchronization and correlation calculation to obtain the original IQ data packet, slice the data packet to obtain the leading code part of each data packet as the initial data for extracting the RF fingerprint feature, and normalize the data to prevent the signal amplitude difference from being too large, which will cause overfitting in the subsequent model training.

[0010] Furthermore, the data collection and preprocessing in step S1 includes the following processes:

[0011] Step S11: Wi-Fi signals are collected from the air by using the USRP-B210 device, and the collected long sequence IQ data is divided into windows. Since the short training period of the Wi-Fi signal has 10 repeated short guide symbols STF at a sampling rate of 20 MHz, and the period is 16 sampling points, the signal in each window is shifted backward by 16 sampling points, and the correlation is calculated with the original signal without shifting to obtain the correlation coefficient matrix for coarse time synchronization. The calculation method is shown in formula (1):

[0012] r short (n) = y(n + L short )y * (n) (1)

[0013] Among them, L short is the period of the short-circuit signal 16, y * (n) represents the conjugate value of the received signal at the nth sampling point.

[0014] Normalize the correlation coefficient matrix, take the threshold as 0.5, and determine the signal starting point by finding the position exceeding the threshold. However, since the short-term symbol includes multiple periodic sequences, there will be multiple starting points with large correlation, so further fine time synchronization is required. Fine time synchronization uses the long-term symbol LTF of the Wi-Fi signal preamble, which has a period of 64 sampling points at a sampling rate of 20MHz. Therefore, the signal in each window is shifted back by 64 sampling points, and the correlation is calculated with the original signal without shifting to obtain the long-period correlation coefficient matrix:

[0015] r long (n) = y(n + L long )y * (n) (2)

[0016] Among them, L long is the period of the long-conducting signal, 64, * (n) represents the conjugate value of the received signal at the nth sampling point.

[0017] Select r short (n) and r long The signal points in the (n) matrix that simultaneously exceed the threshold are taken as the starting points of the Wi-Fi data packets. Since the preamble field of the Wi-Fi signal includes 320 sampling points at a sampling rate of 20 MHz, and there are certain transient characteristics in the process of the signal rising from the initial noise stage to the signal amplitude, a total of 400 signal sampling points, including the 80 bits before the signal starting point and the 320 bits after the starting point, are selected for slicing as subsequent model training and test data.

[0018] Step S12: The data normalization process described in step S1 is mainly to perform data normalization process on the data after the last synchronization and slicing in step S11. The specific calculation method is as shown in formula (3):

[0019]

[0020] Where M represents the length of the signal and is 400.

[0021] Step S2: Based on the normalized preamble signal data in step S1, a data random enhancement method based on physical damage is proposed to perform random frequency and phase enhancement on the training set data for subsequent training models. The enhancement method is:

[0022]

[0023] Among them, formula (4) is used for frequency enhancement of data, formula (5) is used for phase enhancement of data, αK and β represent the frequency and phase size of random enhancement respectively, f k (x b(t),τ) represents the original data expression, Δw k represents the original frequency, represents the original phase, and h(t) represents the modeling expression of the wireless channel.

[0024] Step S3: Based on the enhanced data set in step S2, a method combining the logarithmic spectra of steady-state and transient signals is designed, such as Figure 1 The figure shows a schematic diagram of the structure of the Wi-Fi signal preamble field. The STF signal of the 2nd to 9th cycle and the LTF signal of the 1st to 2nd cycle of the signal preamble part are selected as steady-state features, as shown in formula (6). At the same time, the STF signal of the 1st to 4th cycle and the 9th to 10th STF and GI signals are selected as transient features, as shown in formula (7).

[0025]

[0026]

[0027] Among them, F log (·) represents the operation of taking the log of the signal power spectrum, It represents the first 16 sampling point signal of the first LTF cycle, and concat(·) represents the concatenation operation.

[0028] Finally, the steady-state signal feature RFF Stable And transient signal characteristics RFF Transient The spliced ​​features are used together as robust RF fingerprint features for subsequent model training and testing.

[0029] Step S4: Construct a domain adversarial transfer learning network model. Since the existing RF fingerprint recognition model has poor robustness in complex multi-domain environments, a domain adaptation mechanism is introduced to construct a network model. Figure 2 As shown, it includes a feature extractor module, a label classifier module and a domain classifier module.

[0030] Furthermore, the feature extractor module, the label classifier module and the domain classifier module in step S4 respectively include:

[0031] Step S41: Select the ResNet18 network as the feature extractor module of the model. The specific network structure includes:

[0032] (1) Initialize the convolutional layer

[0033] (2) Residual blocks, including 4 stages, each stage includes 2 residual blocks. Each residual block consists of 2 1*3 convolutional layers and has a skip connection.

[0034] (3) Global average pooling layer, after the last residual block, a global average pooling layer is used to aggregate features.

[0035] Step S42: The label classifier module includes two fully connected layers, a batch normalization layer, a ReLU activation function, and the output channel of the last fully connected layer is the number of label categories.

[0036] Step S43: The domain classifier module includes a gradient reversal layer, two fully connected layers, a batch normalization layer, a ReLU activation function, and the last fully connected output channel is the number of domains, which is set to 2 here, indicating the two domains of the source domain and the target domain. The gradient reversal layer added between the feature extractor and the domain classifier enables the domain classifier to distinguish the data of the source domain and the target domain, while the feature extractor module cannot distinguish whether the data comes from the source domain or the target domain, thereby extracting the common features that are invariant to the domain.

[0037] Step S5: Model training and result evaluation: Use accuracy as the evaluation indicator to train the model and evaluate its signal recognition effect in cross-domain RF fingerprint scenarios. It includes the following processes:

[0038] (1) First, the USRP-B210 device is used to collect Wi-Fi signal data for N days with different channel states.

[0039] (2) The collected data is processed through step S1 to obtain pre-processed signal data.

[0040] (3) Randomly select the data of one day as the training set data, that is, the source domain, and randomly select another day as the test set data, that is, the target domain.

[0041] (4) The training set data is enhanced by the method in step S2 to obtain the enhanced source domain data.

[0042] (5) The source domain data and the target domain data are respectively passed through the model in step S4 to obtain the label classifier output matrix and the domain classifier output matrix.

[0043] (6) The loss function uses cross entropy loss to calculate the classification output loss L of the source domain data. y , domain classifier loss L of source domain data d1 and the domain classifier loss L of the target domain data d2 .

[0044] (7) L y +L d1 +L d2 The gradient update is performed as the total loss of the model.

[0045] Step S6: Model application: Process the source domain data collected in the actual electromagnetic scene and a small portion of the target domain data of the new scene, and obtain the trained model based on steps S1-S5, and then perform the following operations:

[0046] Step S61: Obtain the Wi-Fi signal data to be identified in the actual new scene: collect the Wi-Fi radio frequency signal to be identified through the radio frequency receiving device, extract the domain-invariant radio frequency fingerprint features, and convert them into a model input format.

[0047] Step S62: Use the trained domain adaptation model to perform classification, determine the device identity category corresponding to the signal to be identified, and implement Wi-Fi radio frequency fingerprint identification.

[0048] A second aspect of the present invention relates to a Wi-Fi radio frequency fingerprint recognition device based on domain adaptation, comprising a memory and one or more processors, wherein the memory stores an executable code, and when the one or more processors execute the executable code, they are used to implement a Wi-Fi radio frequency fingerprint recognition method based on domain adaptation of the present invention.

[0049] A third aspect of the present invention relates to a computer-readable storage medium having a program stored thereon, which, when executed by a processor, implements a Wi-Fi radio frequency fingerprint recognition method based on domain adaptation of the present invention.

[0050] The technical concept of the present invention is to propose a data random enhancement method based on physical damage, design a radio frequency fingerprint feature that combines the logarithmic spectrum of the steady-state domain and the transient signal, introduce the domain adaptation mechanism into the field of radio frequency fingerprint recognition, and construct a domain adversarial transfer learning network training model, so that it can realize robust recognition of Wi-Fi radio frequency fingerprints under the condition that the external environment and channel state change.

[0051] The beneficial effects of the present invention are mainly manifested in: through data enhancement and the introduction of a domain adaptation mechanism, radio frequency fingerprint recognition is successfully performed, and compared with traditional machine learning methods, better detection effects are achieved. Experiments show that the method improves the accuracy by about 10%-25% compared with existing models in a single-domain generalization scenario, and solves the problem of data distribution drift of existing models across different temperatures and channel states. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 is a flow chart of the method of the present invention.

[0053] Figure 2 It is the overall framework diagram of the method of the present invention.

[0054] Figure 3 It is a schematic diagram of the structure of the Wi-Fi signal preamble field.

[0055] Figure 4 It is the domain adversarial transfer learning network model structure. DETAILED DESCRIPTION

[0056] To make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with specific embodiments of the present invention and corresponding drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them.

[0057] Example 1

[0058] First combine Figure 1 The present invention provides a Wi-Fi radio frequency fingerprint recognition method according to an embodiment of the present invention. The method comprises the following steps:

[0059] Step S1: Data collection and preprocessing: Construct a Wi-Fi signal detector to collect radio signals from the air, detect the Wi-Fi signal from it through time synchronization and correlation calculation to obtain the original IQ data packet, slice the data packet to obtain the leading code part of each data packet as the initial data for extracting the RF fingerprint feature, and normalize the data to prevent the signal amplitude difference from being too large, which will cause overfitting in the subsequent model training.

[0060] Furthermore, the data collection and preprocessing in step S1 includes the following processes:

[0061] Step S11: Wi-Fi signals are collected from the air by using the USRP-B210 device, and the collected long sequence IQ data is divided into windows. Since the short training period of the Wi-Fi signal has 10 repeated short guide symbols STF at a sampling rate of 20 MHz, and the period is 16 sampling points, the signal in each window is shifted backward by 16 sampling points, and the correlation is calculated with the original signal without shifting to obtain the correlation coefficient matrix for coarse time synchronization. The calculation method is shown in formula (1):

[0062] r short (n) = y(n + L short )y * (n) (1)

[0063] Among them, L short is the period of the short-circuit signal 16, y * (n) represents the conjugate value of the received signal at the nth sampling point.

[0064] Normalize the correlation coefficient matrix, take the threshold as 0.5, and determine the signal starting point by finding the position exceeding the threshold. However, since the short-term symbol includes multiple periodic sequences, there will be multiple starting points with large correlation, so further fine time synchronization is required. Fine time synchronization uses the long-term symbol LTF of the Wi-Fi signal preamble, which has a period of 64 sampling points at a sampling rate of 20MHz. Therefore, the signal in each window is shifted back by 64 sampling points, and the correlation is calculated with the original signal without shifting to obtain the long-period correlation coefficient matrix:

[0065] r long (n) = y(n + L long )y * (n) (2)

[0066] Among them, L long is the period of the long-conducting signal, 64, * (n) represents the conjugate value of the received signal at the nth sampling point.

[0067] Select r short (n) and r long The signal points in the (n) matrix that simultaneously exceed the threshold are taken as the starting points of the Wi-Fi data packets. Since the preamble field of the Wi-Fi signal includes 320 sampling points at a sampling rate of 20 MHz, and there are certain transient characteristics in the process of the signal rising from the initial noise stage to the signal amplitude, a total of 400 signal sampling points, including the 80 bits before the signal starting point and the 320 bits after the starting point, are selected for slicing as subsequent model training and test data.

[0068] Step S12: The data normalization process described in step S1 is mainly to perform data normalization process on the data after the last synchronization and slicing in step S11. The specific calculation method is as shown in formula (3):

[0069]

[0070] Where M represents the length of the signal and is set to 400.

[0071] Specifically, the embodiment based on step S1 can be briefly introduced as follows:

[0072] A USRP-B210 device was used, and the acquisition center frequency was set to 2457 MHz, the bandwidth was set to 20 MHz, and the sampling rate was set to 20 MHz. Wi-Fi signals of 10 mobile devices with different channel states were collected for 9 days as a data set. Each device collected 1,000 training sets and 200 test sets every day, for a total of 108,000 Wi-Fi signal data. Each data was normalized according to the method in step S12 to facilitate subsequent feature extraction.

[0073] Step S2: Based on the normalized preamble signal data in step S1, a data random enhancement method based on physical damage is proposed to perform random frequency and phase enhancement on the training set data for subsequent training models. The enhancement method is:

[0074]

[0075] Among them, formula (4) is used for frequency enhancement of data, formula (5) is used for phase enhancement of data, αK and β represent the frequency and phase size of random enhancement respectively, f k (x b (t),τ) represents the original data expression, Δw k represents the original frequency, represents the original phase, and h(t) represents the modeling expression of the wireless channel.

[0076] Specifically, the embodiment based on step S2 can be briefly introduced as follows:

[0077] Based on the preprocessed data obtained in step S1, the training data is randomly enhanced in frequency deviation and phase deviation according to the method in step S2. The frequency deviation enhancement range can be set to [0, 2000] Hz, and the phase deviation enhancement range can be set to [0, 2] rad to obtain enhanced training data.

[0078] Step S3: Based on the enhanced data set in step S2, a method combining the logarithmic spectra of steady-state and transient signals is designed, such as Figure 1 The figure shows a schematic diagram of the structure of the Wi-Fi signal preamble field. The STF signal of the 2nd to 9th cycle and the LTF signal of the 1st to 2nd cycle of the signal preamble part are selected as steady-state features, as shown in formula (6). At the same time, the STF signal of the 1st to 4th cycle and the 9th to 10th STF and GI signals are selected as transient features, as shown in formula (7).

[0079]

[0080] Among them, F log (·) represents the operation of taking the log of the signal power spectrum, It represents the first 16 sampling point signal of the first LTF cycle, and concat(·) represents the concatenation operation.

[0081] Finally, the steady-state signal feature RFF Stable And transient signal characteristics RFF Transient The spliced ​​features are used together as robust RF fingerprint features for subsequent model training and testing.

[0082] Specifically, the embodiment based on step S3 can be briefly introduced as follows:

[0083] Based on the enhanced data obtained in step S2, the transient features and steady-state features of the domain-invariant RF fingerprint are extracted according to the method in step S3, and the steady-state and transient features are combined and spliced ​​as the input of the subsequent model.

[0084] Step S4: Construct a domain adversarial transfer learning network model. Since the existing RF fingerprint recognition model has poor robustness in complex multi-domain environments, a domain adaptation mechanism is introduced to construct a network model. Figure 2 As shown, it includes a feature extractor module, a label classifier module and a domain classifier module.

[0085] Furthermore, the feature extractor module, the label classifier module and the domain classifier module in step S4 respectively include:

[0086] Step S41: Select the ResNet18 network as the feature extractor module of the model. The specific network structure includes:

[0087] (1) Initialize the convolutional layer

[0088] (2) Residual blocks, including 4 stages, each stage includes 2 residual blocks. Each residual block consists of 2 1*3 convolutional layers and has a skip connection.

[0089] (3) Global average pooling layer, after the last residual block, a global average pooling layer is used to aggregate features.

[0090] Step S42: The label classifier module includes two fully connected layers, a batch normalization layer, a ReLU activation function, and the output channel of the last fully connected layer is the number of label categories.

[0091] Step S43: The domain classifier module includes a gradient reversal layer, two fully connected layers, a batch normalization layer, a ReLU activation function, and the last fully connected output channel is the number of domains, which is set to 2 here, indicating the two domains of the source domain and the target domain. The gradient reversal layer added between the feature extractor and the domain classifier enables the domain classifier to distinguish the data of the source domain and the target domain, while the feature extractor module cannot distinguish whether the data comes from the source domain or the target domain, thereby extracting the common features that are invariant to the domain.

[0092] Step S5: Model training and result evaluation. This includes the following processes:

[0093] (1) First, the USRP-B210 device is used to collect Wi-Fi signal data for N days with different channel states.

[0094] (2) The collected data is processed through step S1 to obtain pre-processed signal data.

[0095] (3) Randomly select the data of one day as the training set data, that is, the source domain, and randomly select another day as the test set data, that is, the target domain.

[0096] (4) The training set data is enhanced by the method in step S2 to obtain the enhanced source domain data.

[0097] (5) The source domain data and the target domain data are respectively passed through the model in step S4 to obtain the label classifier output matrix and the domain classifier output matrix.

[0098] (6) The loss function uses cross entropy loss to calculate the classification output loss L of the source domain data. y , domain classifier loss L of source domain data d1 and the domain classifier loss L of the target domain data d2 .

[0099] (7) L y +L d1 +L d2 The gradient update is performed as the total loss of the model.

[0100] Specifically, the embodiment based on step S5 can be briefly introduced as follows:

[0101] Based on the domain-invariant RF fingerprint features obtained in step S3, according to the training and testing process of step S5, the features are input into the domain adversarial transfer learning network model built in step S4, the learning rate is set to 0.001, the number of training rounds is set to 200 rounds, the optimizer selects the Adam optimizer, and the loss function selects the cross entropy loss for model training.

[0102] Step S6: Model application: Process the source domain data collected in the actual electromagnetic scene and a small portion of the target domain data of the new scene, and obtain the trained model based on steps S1-S5, and then perform the following operations:

[0103] Step S61: Obtain the Wi-Fi signal data to be identified in the actual new scene: collect the Wi-Fi radio frequency signal to be identified through the radio frequency receiving device, extract the domain-invariant radio frequency fingerprint features, and convert them into a model input format.

[0104] Step S62: Use the trained domain adaptation model to perform classification, determine the device identity category corresponding to the signal to be identified, and implement Wi-Fi radio frequency fingerprint identification.

[0105] Specifically, based on the embodiment of step S6, by setting up an experiment, the first day's data is taken as the source domain data for training, and a small amount of data from the remaining 8 days is used as the target domain for testing to achieve single domain generalization, and then new data from the remaining 8 days is collected for model recognition. The test recognition results of the remaining 8 days for the target domain are 94.10%, 93.70%, 94.00%, 89.90%, 96.55%, 95.75%, 86.20%, and 93.70% respectively. At the same time, the experiment is set up by comparing the methods in the existing literature, and the results show that the method of the present invention improves the accuracy by about 10%-25% on average compared with the existing literature methods.

[0106] The present invention proposes a data random enhancement method based on physical damage, designs a radio frequency fingerprint feature that combines the logarithmic spectrum of the steady-state domain and the transient signal, and introduces the domain adaptation mechanism into the field of radio frequency fingerprint recognition. The method aims to improve the cross-domain recognition accuracy of radio frequency fingerprints. The method belongs to the field of signal security technology and successfully realizes the robust cross-domain recognition of Wi-Fi radio frequency fingerprints.

[0107] Example 2

[0108] This embodiment relates to a Wi-Fi radio frequency fingerprint recognition device based on domain adaptation, including a memory and one or more processors, wherein the memory stores an executable code, and when the one or more processors execute the executable code, they are used to implement a Wi-Fi radio frequency fingerprint recognition method based on domain adaptation in embodiment 1.

[0109] Example 3

[0110] This embodiment relates to a computer-readable storage medium on which a program is stored. When the program is executed by a processor, a Wi-Fi radio frequency fingerprint recognition method based on domain adaptation of the present invention is implemented.

[0111] The contents described in the embodiments of this specification are merely an enumeration of the implementation forms of the inventive concept. The protection scope of the present invention should not be regarded as limited to the specific forms described in the embodiments. The protection scope of the present invention also extends to equivalent technical means that can be conceived by those skilled in the art based on the inventive concept.

Claims

1. A Wi-Fi radio frequency fingerprint recognition method based on domain adaptation, characterized in that: The following steps are involved: Step S1: Data collection and preprocessing: Construct a Wi-Fi signal detector to collect radio signals from the air, detect the Wi-Fi signal from it through time synchronization and correlation calculation to obtain the original IQ data packet, slice the data packet to obtain the leading code part of each data packet as the initial data for extracting the RF fingerprint feature, and normalize the data to prevent the large difference in signal amplitude from causing overfitting in subsequent model training; Step S2: Based on the normalized preamble signal data in step S1, a data random enhancement method based on physical damage is proposed to perform random frequency and phase enhancement on the training set data for subsequent training models; Step S3: Based on the enhanced data set in step S2, a method combining the logarithmic spectra of steady-state and transient signals is designed, and the STF signal of the 2nd to 9th period and the LTF signal of the 1st to 2nd period of the signal preamble part are selected as steady-state features, and the STF signal of the 1st to 4th period and the 9th to 10th STF and GI signal of the signal are selected as transient features; Then the steady-state signal feature RFF Stable And transient signal characteristics RFF Transient The two are concatenated together as robust RF fingerprint features for subsequent model training and testing; Step S4: constructing a domain adversarial transfer learning network model. Since the existing radio frequency fingerprint recognition model has poor robustness in a complex multi-domain environment, a domain adaptation mechanism is introduced to construct a network model; the network model includes a feature extractor module, a label classifier module and a domain classifier module; Step S5: Model training and result evaluation process: Use accuracy as the evaluation indicator to train the model and evaluate its signal recognition effect in cross-domain RF fingerprint scenarios; Step S6: Model application: Process the source domain data collected in the actual electromagnetic scene and a small portion of the target domain data of the new scene, and obtain the trained model based on steps S1-S5, and then perform the following operations: Step S61: Obtaining Wi-Fi signal data to be identified in an actual new scenario: collecting the Wi-Fi radio frequency signal to be identified through a radio frequency receiving device, extracting domain-invariant radio frequency fingerprint features, and converting them into a model input format; Step S62: Use the trained domain adaptation model to perform classification, determine the device identity category corresponding to the signal to be identified, and implement Wi-Fi radio frequency fingerprint identification.

2. A domain adaptive Wi-Fi radio frequency fingerprint recognition method according to claim 1, characterized in that: The data collection and preprocessing in step S1 includes the following processes: Step S11: Wi-Fi signals are collected from the air by using the USRP-B210 device, and the collected long sequence IQ data is divided into windows. Since the short training period of the Wi-Fi signal has 10 repeated short guide symbols STF at a sampling rate of 20 MHz, and the period is 16 sampling points, the signal in each window is shifted backward by 16 sampling points, and the correlation is calculated with the original signal without shifting to obtain the correlation coefficient matrix for coarse time synchronization. The calculation method is shown in formula (1): r short (n)=y(n+L short )y * (n) (1) Among them, L short is the period of the short-circuit signal 16, y * (n) represents the conjugate value of the received signal at the nth sampling point; Normalize the correlation coefficient matrix, take the threshold as 0.5, and determine the signal starting point by finding the position exceeding the threshold; however, since the short guide symbol includes multiple periodic sequences, there will be multiple starting points with large correlation, so further fine time synchronization is required; fine time synchronization uses the long guide symbol LTF of the Wi-Fi signal preamble, and the period at the 20MHz sampling rate is 64 sampling points. Therefore, the signal in each window is shifted back by 64 sampling points, and the correlation is calculated with the original signal without shifting, and the long period correlation coefficient matrix is ​​obtained: r long (n)=y(n+L long )y * (n) (2) Among them, L long is the period of the long-conducting signal, 64, * (n) represents the conjugate value of the received signal at the nth sampling point; Select r short (n) and r long (n) The signal points in the matrix that simultaneously exceed the threshold are taken as the starting points of the Wi-Fi data packets. Since the preamble field of the Wi-Fi signal includes 320 sampling points at a sampling rate of 20 MHz, and there are certain transient characteristics in the process of the signal rising from the initial noise stage to the signal amplitude, a total of 400 signal sampling points, including 80 bits before the signal starting point and 320 bits after the starting point, are selected for slicing as subsequent model training and test data; Step S12: The data normalization process described in step S1 is to perform data normalization process on the data after the last synchronization and slicing in step S11. The specific calculation method is as shown in formula (3): Where M represents the length of the signal and is set to 400.

3. A domain adaptive Wi-Fi radio frequency fingerprint recognition method according to claim 1, characterized in that: Step S2 specifically includes: Based on the normalized preamble signal data in step S1, a data random enhancement method based on physical damage is proposed to perform random frequency and phase enhancement on the training set data for subsequent training models; the enhancement method is: Among them, formula (4) is used for frequency enhancement of data, formula (5) is used for phase enhancement of data, αK and β represent the frequency and phase size of random enhancement respectively, f k (x b (t),τ) represents the original data expression, Δw k represents the original frequency, represents the original phase, and h(t) represents the modeling expression of the wireless channel.

4. The Wi-Fi radio frequency fingerprint recognition method based on domain adaptation according to claim 1, characterized in that: In step S3, the STF signal of the 2nd to 9th cycle and the LTF signal of the 1st to 2nd cycle of the preamble code part of the selected signal are used as steady-state features, and formula (6) is used. The STF signal of the 1st to 4th cycle and the 9th to 10th STF and GI signal of the selected signal are used as transient features, and formula (7) is used: Among them, F log (·) represents the operation of taking the log of the signal power spectrum, It represents the first 16 sampling point signal of the first LTF cycle, and concat(·) represents the concatenation operation.

5. The Wi-Fi radio frequency fingerprint recognition method based on domain adaptation according to claim 1, characterized in that: Step S4 specifically includes: Construct a domain adversarial transfer learning network model. Since the existing RF fingerprint recognition model has poor robustness in complex multi-domain environments, a domain adaptation mechanism is introduced to construct a network model; including a feature extractor module, a label classifier module, and a domain classifier module; In step S4, the ResNet18 network is selected as the feature extractor module of the model; the specific network structure includes: (1) Initialize the convolutional layer (2) Residual blocks, including 4 stages, each stage includes 2 residual blocks; each residual block consists of 2 1*3 convolutional layers and has a skip connection; (3) Global average pooling layer: After the last residual block, a global average pooling layer is used to aggregate features; The label classifier module described in step S4 includes two fully connected layers, a batch normalization layer, a ReLU activation function, and the output channel of the last fully connected layer is the number of label categories; The domain classifier module described in step S4 includes a gradient reversal layer, two fully connected layers, a batch normalization layer, a ReLU activation function, and the last fully connected output channel is the number of domains, which is set to 2 here, indicating the two domains of the source domain and the target domain. The gradient reversal layer added between the feature extractor and the domain classifier enables the domain classifier to distinguish the data of the source domain and the target domain, while the feature extractor module cannot distinguish whether the data comes from the source domain or the target domain, thereby extracting the common features that are invariant to the domain.

6. A domain adaptive Wi-Fi radio frequency fingerprint recognition method according to claim 1, characterized in that: Step S5 specifically includes: (1) First, use the USRP-B210 device to collect Wi-Fi signal data for N days with different channel states; (2) The collected data is processed through step S1 to obtain pre-processed signal data; (3) Randomly select one day’s data as the training set data, i.e., the source domain, and randomly select another day as the test set data, i.e., the target domain; (4) The training set data is enhanced by the method in step S2 to obtain the enhanced source domain data. (5) The source domain data and the target domain data are respectively passed through the model in step S4 to obtain a label classifier output matrix and a domain classifier output matrix; (6) The loss function uses cross entropy loss to calculate the classification output loss L of the source domain data. y , domain classifier loss L of source domain data d1 and the domain classifier loss L of the target domain data d2 ; (7) L y +L d1 +L d2 The gradient update is performed as the total loss of the model.

7. A Wi-Fi radio frequency fingerprint recognition device based on domain adaptation, characterized in that: The method comprises a memory and one or more processors, wherein the memory stores executable codes, and when the one or more processors execute the executable codes, the method is used to implement a domain adaptive Wi-Fi radio frequency fingerprint recognition method according to any one of claims 1 to 6.

8. A computer-readable storage medium, characterized in that: A program is stored thereon, and when the program is executed by a processor, a Wi-Fi radio frequency fingerprint recognition method based on domain adaptation according to any one of claims 1 to 6 is implemented.

Images