A dynamic physical layer authentication method and system based on a compensation mechanism

Abstract

The application relates to a dynamic physical layer authentication method and system based on a compensation mechanism, and belongs to the field of wireless communication security. The method utilizes physical layer characteristic information, dynamically predicts the physical layer characteristics of equipment by introducing a Kalman filter, and combines a Gaussian kernel function to calculate the similarity between actual measured characteristic information and predicted characteristic information, so that accurate identification of the equipment identity is realized. The application designs a physical layer authentication model based on error feedback, which is used for real-time adjustment of kernel function parameters and compensation of characteristic deviation caused by factors such as environmental changes and channel disturbance. By combining the physical layer characteristics with the kernel function model, the application realizes a lightweight authentication mechanism without key distribution and complex password calculation, effectively reduces authentication delay and system resource overhead. The application has dynamic updating capability and strong anti-attack performance, and can be widely applied to security authentication scenes such as Internet of Things terminals, edge devices and wireless communication networks.

Description

TECHNICAL FIELD

[0001] The present application belongs to the field of wireless communication security, and particularly relates to a dynamic physical layer authentication method and system based on a compensation mechanism, which is suitable for wireless Internet of Things, edge device access and the like. BACKGROUND

[0002] With the rapid development of wireless communication technology and the wide deployment of Internet of Things devices, the traditional identity authentication mechanism based on cryptography gradually exposes the limitations in the resource-constrained environment. For example, in edge nodes, embedded terminals and the like, the device computing capability is limited and cannot bear complex key negotiation, encryption and decryption and the like. In addition, the key management system itself also faces the risk of being stolen, leaked or synchronization failure, resulting in reduced overall system security.

[0003] In recent years, researchers have begun to focus on physical layer authentication technology, the core idea of which is to model and identify the "hardware fingerprint" of the device by using the instantaneous characteristics in the wireless channel that are difficult to forge, such as carrier frequency offset, received signal strength, channel state information and the like. Since these characteristics are subject to spatial position, hardware differences and environmental disturbances, they have a natural unclonability to some extent, and are widely used in lightweight security authentication.

[0004] However, the existing physical layer-based authentication methods still have many challenges. First, the physical layer characteristics are highly sensitive to the environment, and changes in indoor layout or personnel movement can cause characteristic shifts, thereby affecting the authentication accuracy. Second, most algorithms use fixed thresholds or static models, which lack adaptability and are difficult to cope with dynamic and complex wireless environments. In addition, the feature matching process usually does not consider continuity and historical state, and is easily bypassed by attackers through replay or relay attacks. Therefore, there is an urgent need for a lightweight physical layer authentication mechanism that can both dynamically compensate for feature changes and adaptively optimize the decision-making process. SUMMARY

[0005] The purpose of the present application is to overcome the defects of lack of adaptability, poor robustness and susceptibility to environmental interference in the existing physical layer authentication system, and to provide a dynamic physical layer authentication method and system based on a compensation mechanism, which can effectively improve the stability and security of physical layer authentication in actual wireless environments by introducing a dynamic adjustment mechanism.

[0006] To achieve the above purpose, the technical scheme of the present application is: a dynamic physical layer authentication method based on a compensation mechanism, comprising:

[0007] S1, a feature extraction stage: from the received wireless signal, physical layer features including at least carrier frequency offset CFO, received signal strength RSSI and channel state information CSI are extracted to form a current time feature value vector , is the first moment, and performs normalization processing;

[0008] S2, trust phase:

[0009] S21, eigenvalue vector prediction: the eigenvalue is predicted and updated by the Kalman filter to generate the current eigenvalue prediction vector , and performs normalization processing;

[0010] S22, similarity calculation: the similarity value between the current eigenvalue vector and the current eigenvalue prediction vector is calculated using the Gaussian kernel function ;

[0011] S23, error calculation and dynamic compensation mechanism: the current error term is calculated, and based on the current error term , the physical layer authentication model based on error feedback is used to dynamically update the kernel function bandwidth parameter at the current moment;

[0012] S24, repeat step S1, step S21 to step S23 until the kernel function bandwidth parameter is less than the preset value, and enter the authentication phase;

[0013] S3, authentication phase:

[0014] S31, eigenvalue vector prediction: the eigenvalue is predicted by the Kalman filter to generate the current eigenvalue prediction vector ;

[0015] S32, authentication decision: the similarity value is calculated according to the current eigenvalue vector , the current eigenvalue prediction vector and the updated using the Gaussian kernel function, if is greater than the authentication threshold , the authentication is successful; otherwise, the authentication fails and access is rejected;

[0016] S33, parameter update: if the authentication is successful, the kernel function bandwidth parameter is updated and the update step of the Kalman filter is executed; otherwise, the kernel function bandwidth parameter is maintained and the state of the Kalman filter is unchanged;

[0017] S34, threshold update: based on the latest authentication result, the threshold is dynamically adjusted;

[0018] S35, repeat step S1, step S31 to step S34 until the end of the communication.

[0019] Further, based on the current time error term , the calculation formula of the kernel function bandwidth parameter of the current time is dynamically updated by using the physical layer authentication model based on error feedback The calculation formula is as follows:

[0020]

[0021] Among them, is the kernel function bandwidth parameter of the next time, is the error term of the last time, is the error term of the last two times, is the proportional gain, is the integral gain, is the differential gain, which is used to dynamically control the change of the kernel function bandwidth parameter, so that the system can maintain the stability and response ability of the authentication model in a variable wireless environment.

[0022] Further, the Gaussian kernel function is in the form of:

[0023] .

[0024] Further, the Gaussian kernel function can be replaced by other linear similarity calculation functions.

[0025] Further, the initial Kalman filter state and the initial kernel function bandwidth parameter value are obtained in the trust stage, and the trust stage is trained by a known legal device for several rounds to build a basic model.

[0026] Further, and the authentication threshold The comparison formula is as follows:

[0027]

[0028] Indicates that this authentication is passed, and the corresponding device is a legal device; Indicates that this authentication fails, and the corresponding device is an illegal device.

[0029] The application also provides a dynamic physical layer authentication system based on a compensation mechanism, comprising a signal receiving module, a physical layer feature extraction module, a Kalman filter prediction module, a Gaussian kernel similarity calculation module, a physical layer authentication model based on error feedback, and an adaptive threshold adjustment module; The authentication system performs the following steps:

[0030] Trust stage:

[0031] (1) Receive wireless signals through the signal receiving module, and extract physical layer features from the wireless signals through the physical layer feature extraction module, including at least carrier frequency offset (CFO), received signal strength (RSSI), and channel state information (CSI), to form the feature value vector at the current time. , For the first The time frame was then normalized.

[0032] (2) Eigenvalue vector prediction: The eigenvalues ​​are predicted and updated by the Kalman filter prediction module to generate the eigenvalue prediction vector at the current time. and normalize it;

[0033] (3) Similarity calculation: The Gaussian kernel similarity calculation module is used to calculate the feature vector at the current time. With the current time-time eigenvalue prediction vector Similarity values ​​between ;

[0034] (4) Error calculation and dynamic compensation mechanism: calculate the error term at the current time. Based on the error term at the current time The kernel function bandwidth parameter at the current moment is dynamically updated using a physical layer authentication model based on error feedback. ;

[0035] (5) Repeat steps (1) to (4) until the kernel function bandwidth parameter is reached. If the value is less than the preset value, proceed to the authentication stage;

[0036] Certification phase:

[0037] (1) Receive wireless signals through the signal receiving module, and extract physical layer features from the wireless signals through the physical layer feature extraction module, including at least carrier frequency offset (CFO), received signal strength (RSSI), and channel state information (CSI), to form the feature value vector at the current time. , For the first The time frame was then normalized.

[0038] (2) Eigenvalue vector prediction: The eigenvalues ​​are predicted by the Kalman filter prediction module to generate the eigenvalue prediction vector at the current time. ;

[0039] (3) Authentication decision: The Gaussian kernel similarity calculation module is used to calculate the authentication result based on the feature vector at the current time. With the current time-time eigenvalue prediction vector And the updated Calculate similarity value ,like greater than the authentication threshold authentication succeeds; otherwise, authentication fails, and access is denied;

[0040] (4) Parameter update: if authentication succeeds, update the kernel function bandwidth parameter through the physical layer authentication model based on error feedback and the adaptive threshold adjustment module and perform the update step of the Kalman filter to dynamically adjust the threshold ; otherwise, keep the kernel function bandwidth parameter , and the state of the Kalman filter remains unchanged.

[0041] (5) Enter the judgment of the next frame of signal until the communication ends.

[0042] Further, based on the error term at the current time , the kernel function bandwidth parameter at the current time is dynamically updated using the physical layer authentication model based on error feedback The calculation formula is as follows:

[0043]

[0044] wherein, is the kernel function bandwidth parameter at the next time, is the error term at the last time, is the error term at the last two times, is the proportional gain, is the integral gain, is the derivative gain, which is used to dynamically control the change of the kernel function bandwidth parameter, so that the system maintains the stability and response ability of the authentication model in a variable wireless environment.

[0045] Further, the Gaussian kernel function has the form:

[0046] .

[0047] Further, the Gaussian kernel function can be replaced by other linear similarity calculation functions.

[0048] Further, the initial Kalman filter state and the initial kernel function bandwidth parameter value are obtained in the trust phase, and the trust phase is trained by a known legal device for several rounds to build a basic model.

[0049] Further, and the authentication threshold The comparison formula is as follows:

[0050]

[0051] indicates that this authentication is passed, and the corresponding device is a legal device; indicates that this authentication fails, and the corresponding device is an illegal device.

[0052] Further, the Kalman filter prediction module uses state transition model and noise estimation to filter the measurement error when processing the features, thereby enhancing the predictability and continuity of the feature values.

[0053] Further, when a new device joins the system, the device first enters the trust phase for initialization modeling to obtain the corresponding , and then enters the authentication phase to compare and authenticate the physical layer features with the existing devices in the system.

[0054] Further, the system supports a dynamic learning mechanism, which automatically updates the feature value prediction vector and the physical layer authentication model parameters based on error feedback according to the authentication result of the previous step in the continuous authentication process, and adaptively adjusts the authentication threshold according to the recent authentication result, thereby improving the identity verification accuracy in the continuous communication process.

[0055] Compared with the prior art, the present application has the following beneficial effects: the present application introduces an error feedback-based physical layer authentication model to dynamically adjust the kernel function bandwidth parameter , so that the system can compensate for the physical layer feature deviation caused by environmental changes, hardware drift or dynamic interference, thereby significantly improving the robustness and stability of the authentication system. Compared with the traditional physical layer authentication method using fixed threshold or static model, the present application can flexibly adjust the model parameters according to the actual error feedback, thereby effectively reducing the false reception rate and rejection rate. In addition, the present application uses a Kalman filter to predict and update the physical layer features, thereby further enhancing the tolerance of the system to noise and short-term fluctuations. Experimental results show that the authentication accuracy of the present application in a complex wireless environment is significantly better than that of the prior art, and the authentication success rate and attack resistance performance have been significantly improved, and the present application is particularly suitable for application scenarios such as Internet of Things, edge computing and secure communication, which require low latency and high reliability. BRIEF DESCRIPTION OF DRAWINGS

[0056] Figure 1 is a flowchart of the dynamic physical layer authentication method based on the compensation mechanism of the embodiments of the present application;

[0057] Figure 2 is a schematic diagram of the kernel function bandwidth parameter of the dynamic physical layer authentication method based on the compensation mechanism of the embodiments of the present application in the experiment;

[0058] Figure 3 is a ROC curve schematic diagram in the experiment of the dynamic physical layer authentication method based on the compensation mechanism of the embodiments of the present application;

[0059] Figure 4 This is a schematic diagram illustrating the experimental results of the dynamic physical layer authentication method based on the compensation mechanism in this application. Detailed Implementation

[0060] The above-described solution will be further illustrated below with reference to specific embodiments. It should be understood that these embodiments are for illustrative purposes only and are not intended to limit the scope of this application. The implementation conditions used in the embodiments may be further adjusted according to the conditions of specific manufacturers, and the implementation conditions not specified are generally those in routine experiments.

[0061] This invention provides a dynamic physical layer authentication method based on a compensation mechanism, comprising:

[0062] S1. Feature Extraction Stage: Extract physical layer features from the received wireless signal, including at least carrier frequency offset (CFO), received signal strength (RSSI), and channel state information (CSI), to form the feature value vector at the current time. , For the first The time frame was then normalized.

[0063] S2, Trust Phase:

[0064] S21. Eigenvalue Vector Prediction: The eigenvalues ​​are predicted and updated using a Kalman filter to generate the eigenvalue prediction vector for the current time step. and normalize it;

[0065] S22. Similarity Calculation: The Gaussian kernel function is used to calculate the feature vector at the current time step. With the current time-time eigenvalue prediction vector Similarity value between ;

[0066] S23. Error Calculation and Dynamic Compensation Mechanism: Calculate the error term at the current moment. Based on the error term at the current time The kernel function bandwidth parameter at the current moment is dynamically updated using a physical layer authentication model based on error feedback. ;

[0067] S24. Repeat steps S1, S21 to S23 until the kernel function bandwidth parameter is reached. If the value is less than the preset value, proceed to the authentication stage;

[0068] S3, Certification Phase:

[0069] S31. Eigenvalue Vector Prediction: Eigenvalues ​​are predicted using a Kalman filter to generate the eigenvalue prediction vector for the current time step. ;

[0070] S32, authentication decision: using Gaussian kernel function, according to the current time characteristic value vector and the current time characteristic value prediction vector and the updated Calculate the similarity value , if is greater than the authentication threshold , the authentication is successful, otherwise the authentication fails and access is refused;

[0071] S33, parameter update: if the authentication is successful, update the kernel function bandwidth parameter and execute the update step of Kalman filter; otherwise, keep the kernel function bandwidth parameter , and the state of Kalman filter is unchanged;

[0072] S34, threshold update: based on the latest authentication result, dynamically adjust the threshold ;

[0073] S35, repeat step S1, step S31 to step S34 until the communication ends.

[0074] The application also provides a dynamic physical layer authentication system based on compensation mechanism, comprising a signal receiving module, a physical layer feature extraction module, a Kalman filter prediction module, a Gaussian kernel similarity calculation module, a physical layer authentication model based on error feedback and an adaptive threshold adjustment module; the authentication system executes the following steps:

[0075] Trust phase:

[0076] (1) receiving wireless signals through the signal receiving module, and extracting physical layer features including at least carrier frequency offset CFO, received signal strength RSSI and channel state information CSI from the wireless signals through the physical layer feature extraction module, to form the current time characteristic value vector , is the first time, and normalized processing is performed;

[0077] (2) characteristic value vector prediction: predicting and updating the characteristic value through the Kalman filter prediction module to generate the current time characteristic value prediction vector , and normalized processing is performed;

[0078] (3) similarity calculation: using the Gaussian kernel similarity calculation module to calculate the similarity value between the current time characteristic value vector and the current time characteristic value prediction vector ;

[0079] (4) error calculation and dynamic compensation mechanism: calculating the current time error term , based on the current time error term , a physical layer authentication model based on error feedback and an adaptive threshold adjustment module are used to dynamically update the kernel function bandwidth parameter of the current time ;

[0080] (5) repeat steps (1) to (4) until the kernel function bandwidth parameter is less than the preset value, and enter the authentication phase;

[0081] Authentication phase:

[0082] (1) receiving a wireless signal through a signal receiving module, and extracting at least physical layer features including carrier frequency offset CFO, received signal strength RSSI and channel state information CSI from the wireless signal through a physical layer feature extraction module to form a current time feature value vector , is the first time, and performing normalization processing;

[0083] (2) feature value vector prediction: predicting the feature value through a Kalman filter prediction module to generate a current time feature value prediction vector ;

[0084] (3) authentication decision: using a Gaussian kernel similarity calculation module to calculate a similarity value based on the current time feature value vector , the current time feature value prediction vector and the updated , and if is greater than an authentication threshold , the authentication is successful; otherwise, the authentication fails and access is denied;

[0085] (4) parameter update: if the authentication is successful, updating the kernel function bandwidth parameter through a physical layer authentication model based on error feedback and an adaptive threshold adjustment module, and performing an update step of the Kalman filter to dynamically adjust the threshold ; otherwise, keeping the kernel function bandwidth parameter unchanged, and the state of the Kalman filter is unchanged;

[0086] (5) entering the judgment of the next frame signal until the communication ends.

[0087] The application discloses a dynamic physical layer authentication method based on a compensation mechanism.

[0088] Next, the dynamic authentication method proposed in the application is described in combination with the drawings.

[0089] As Figure 1 shown is a dynamic physical layer authentication method based on a compensation mechanism, which comprises the following steps:

[0090] 1) signal receiving and physical layer feature extraction: the system receives a wireless signal sent by a terminal device, extracts key physical layer feature values such as carrier frequency offset, received signal strength indication, channel state information and the like, and forms a multi-dimensional feature value vector , is the feature value vector at the first time, and the feature value vector is normalized.

[0091] 2) trust stage: in the initialization stage of communication connection, the trust stage is entered:

[0092] 2-1 the system adopts a Kalman filter to predict and update continuously collected physical layer features, and generates a feature value prediction vector .

[0093] 2-2 the similarity between the current feature and the feature value prediction vector is calculated through a Gaussian kernel function:

[0094]

[0095] wherein is the feature value vector at the current time, is the feature value prediction vector at the current time, is the kernel function bandwidth parameter at the current time, and represents the similarity output.

[0096] 2-3 an error signal is constructed:

[0097]

[0098] The error is input as the difference between the system target and the current authentication state.

[0099] 2-4 the kernel function bandwidth parameter is dynamically adjusted through a physical layer authentication model based on error feedback​ value to achieve real-time compensation for channel state changes.

[0100] 2-5 Repeat steps 1) and 4 steps 2-1 to 2-4 until the authentication phase.

[0101] 3) Authentication phase: in the process of normal communication of the system, enter the authentication phase:

[0102] 3-1 The system uses Kalman filter to predict the continuously collected physical layer features, and generates a feature value prediction vector .

[0103] 3-2 Calculate the similarity between the current feature value vector and the feature value prediction vector , get the similarity .

[0104] 3-3 The system compares the similarity value with the set threshold :

[0105] If , determine that the terminal is a legal device, allow access, and update the Kalman filter state, the kernel function bandwidth parameter and the authentication threshold .

[0106] If , it is determined that it is an illegal device, access is denied, and no update step is performed.

[0107] 3-4 Repeat steps 1) and 3-1 to 3-3 three steps until the communication ends.

[0108] Next, a computer simulation experiment is used to realize the dynamic physical layer authentication method based on the compensation mechanism.

[0109] A, experimental conditions:

[0110] In the office environment , Bob (the receiving end) is located in the center of the office; Alice (the sending end one) moves in the lower left area of Bob at a speed of 0.2 m / s, and the signal transmission power is 20 dBm; Eve (the sending end two) moves in the upper right area of Bob at a speed of 0.1 m / s, and the signal transmission power is 15 dBm;

[0111] B, experimental steps:

[0112] The following process is used to realize the dynamic physical layer authentication based on the compensation mechanism.

[0113] 1) Alice and Bob establish a connection.

[0114] 2) Trust phase: Bob extracts CFO, RSSI and CSI amplitude from each received frame signal as a feature value vector . The feature value prediction vector is obtained by prediction and update of the Kalman filter, and the similarity between the feature value vector and the feature value prediction vector is calculated by the Gaussian kernel function. Then, the difference between the similarity and the target similarity 0.999 is input into the physical layer authentication model based on error feedback to adaptively adjust the kernel function bandwidth parameter . These steps are repeated, and finally the kernel function bandwidth parameter is stabilized at a small value.

[0115] 3) Authentication phase: Eve sends attack information to Bob at irregular intervals, Bob extracts a feature value vector from each received frame signal, and a feature value prediction vector is obtained by prediction. The similarity between the feature value vector and the feature value prediction vector is calculated by the Gaussian kernel function, and it is determined whether it is greater than the authentication threshold . If it is greater, the authentication is passed, the difference between the similarity and the target similarity 0.999 is input into the physical layer authentication model based on error feedback to update the kernel function bandwidth parameter and perform the update step of the Kalman filter and the threshold update step; if it is less, the authentication fails, and no update operation is performed, and the next frame signal is judged.

[0116] C, experimental results:

[0117] As shown in Figure 2 , the kernel function bandwidth parameter changes dynamically over time. It can be seen that the system quickly adjusts the kernel function bandwidth parameter in the initial stage, and then the kernel function bandwidth parameter value converges to a stable interval, indicating that the physical layer authentication model based on error feedback proposed in the present application can adaptively adjust the sensitivity of the authentication model according to error feedback, thereby improving the robustness of the authentication model to channel disturbance. The entire authentication process does not require additional keys or manual intervention, and has good adaptive ability and deployment flexibility.

[0118] Figure 3As shown, the ROC curve of the identity authentication model of the present application on the test set is shown. The horizontal axis represents the false acceptance rate, and the vertical axis represents the correct recognition rate. Figure 3 The dashed line represents the performance baseline of the random classifier, and the curve represents the actual performance of the authentication model of the present application. As can be seen, the ROC curve of the present application is close to the upper left corner as a whole, indicating that it has high recognition ability and meets the accuracy requirements of identity authentication in the actual edge collaboration environment.

[0119] Figure 4 As shown, in the dynamic physical layer authentication method based on the compensation mechanism disclosed in the example of the present application, the similarity of different devices in the entire authentication process has obvious differences. In the experiment, the similarity of the legitimate user Alice is always higher than the dynamic threshold, while the similarity of the attacker is mostly below the threshold, and the authentication system can effectively distinguish between legitimate devices and illegal intruders, indicating that the dynamic threshold mechanism proposed by the present application has good discrimination ability. In the actual wireless environment, even if there are complex channel fluctuations, this method can still maintain high authentication stability and security.

[0120] In summary, the authentication method proposed by the present application can achieve accurate recognition of device identity without increasing additional communication overhead and computational burden, and improve the security and real-time performance of authentication. The method is suitable for distributed authentication scenarios of large-scale devices in the Internet of Things environment, and can be widely applied in the fields of edge computing, wireless communication and low-power network, etc. The above embodiments are only used to illustrate the technical concept and advantages of the present application, and any equivalent replacement or functional conversion made based on the spirit of the present application should be covered within the protection scope of the present application.

[0121] The above is the preferred embodiment of the present application, and any changes made based on the technical solution of the present application, which do not exceed the scope of the technical solution of the present application, are within the protection scope of the present application. The above is the preferred embodiment of the present application, and any changes made based on the technical solution of the present application, which do not exceed the scope of the technical solution of the present application, are within the protection scope of the present application.

Claims

1. A dynamic physical layer authentication method based on a compensation mechanism, characterized in that, Comprise: S1, feature extraction stage: from the received wireless signal, at least physical layer features including carrier frequency offset CFO, received signal strength RSSI and channel state information CSI are extracted to form a current time feature value vector , for the first time, and normalized processing is performed; S2, trust phase: S21, eigenvalue vector prediction: the eigenvalue is predicted and updated by a Kalman filter to generate the eigenvalue prediction vector at the current time and normalized processing; S22, similarity calculation: using a Gaussian kernel function to calculate the similarity value between the current time characteristic value vector and the current time characteristic value prediction vector ;​​ S23, error calculation and dynamic compensation mechanism: calculate the error term at the current time , based on the error term at the current time , dynamically update the kernel function bandwidth parameter at the current time using the physical layer authentication model based on error feedback ; S24, repeat step S1, step S21 to step S23 until the kernel function bandwidth parameter less than a preset value, enter the authentication phase; S3, authentication phase: S31, eigenvalue vector prediction: the eigenvalue is predicted by a Kalman filter to generate an eigenvalue prediction vector at the current time ; S32, authentication decision: using Gaussian kernel function, according to the current time characteristic value vector and the current time characteristic value prediction vector and the updated Calculate the similarity value If is greater than the authentication threshold , the authentication is successful; otherwise, the authentication fails and access is refused; S33, parameter updating: if the authentication is successful, update the kernel function bandwidth parameter and perform the updating step of the Kalman filter; otherwise, keep the kernel function bandwidth parameter unchanged, and the state of the Kalman filter is unchanged S34, threshold updating: dynamically adjust the threshold based on the recent authentication results ; S35, repeat step S1, step S31 to step S34 until the end of communication.

2. The dynamic physical layer authentication method based on compensation mechanism according to claim 1, characterized in that, Based on the current time error term , the physical layer authentication model based on error feedback is adopted to dynamically update the kernel function bandwidth parameter of the current time The calculation formula is as follows: wherein, is the kernel bandwidth parameter for the next time instant, is the error term for the previous time instant, is the error term for the two previous time instants, is the proportional gain, is the integral gain, is the derivative gain, used to dynamically control the variation of the kernel bandwidth parameter, so that the system maintains the stability and response ability of the authentication model in a variable wireless environment.

3. The dynamic physical layer authentication method based on compensation mechanism according to claim 1, characterized in that, The form of the Gaussian kernel function is: 。 4. The dynamic physical layer authentication method based on compensation mechanism according to claim 1, characterized in that, In the trust phase, the initial Kalman filter state and the initial kernel function bandwidth parameter value are obtained, and the trust phase is trained by a known legal device for several rounds to build a basic model.

5. The dynamic physical layer authentication method based on compensation mechanism according to claim 1, characterized in that, with the authentication threshold The comparison formula is expressed as follows: indicates that the current authentication is passed, and the corresponding device is a legal device; indicates that the current authentication is failed, and the corresponding device is an illegal device.

6. A dynamic physical layer authentication system based on a compensation mechanism, characterized in that, Comprise signal receiving module, physical layer feature extraction module, Kalman filter prediction module, Gaussian kernel similarity calculation module, physical layer authentication model based on error feedback and adaptive threshold adjustment module; the authentication system executes the following steps: Trust phase: (1) receiving a wireless signal through a signal receiving module, and extracting at least physical layer features including carrier frequency offset CFO, received signal strength RSSI and channel state information CSI from the wireless signal through a physical layer feature extracting module to form a feature value vector at a current time , is the first time, and normalization processing is performed; (2) Eigenvalue vector prediction: the eigenvalue is predicted and updated by a Kalman filter prediction module to generate an eigenvalue prediction vector at the current time and normalized processing; (3) Similarity calculation: a Gaussian kernel similarity calculation module is used to calculate the similarity value between the current time characteristic value vector and the current time characteristic value prediction vector ;​ (4) Error calculation and dynamic compensation mechanism: calculate the error term at the current time , based on the error term at the current time , the physical layer authentication model based on error feedback is used to dynamically update the kernel function bandwidth parameter at the current time ; (5) repeat steps (1) to (4) until the kernel function bandwidth parameter is less than a preset value, entering an authentication phase; Authentication phase: (1) receiving a wireless signal through a signal receiving module, and extracting at least physical layer features including carrier frequency offset CFO, received signal strength RSSI and channel state information CSI from the wireless signal through a physical layer feature extracting module to form a feature value vector at a current time , is the first time, and normalization processing is performed; (2) Eigenvalue vector prediction: the eigenvalue is predicted by a Kalman filter prediction module to generate an eigenvalue prediction vector at the current time ; (3) authentication decision: using a Gaussian kernel similarity calculation module, according to the current time characteristic value vector and the current time characteristic value prediction vector and the updated , calculate the similarity value , if is greater than the authentication threshold , the authentication is successful; otherwise, the authentication fails and access is refused; (4) Parameter updating: if the authentication is successful, update the kernel function bandwidth parameter through the physical layer authentication model based on error feedback and the adaptive threshold adjustment module and perform the updating step of the Kalman filter to dynamically adjust the threshold ; otherwise, keep the kernel function bandwidth parameter , and the state of the Kalman filter remains unchanged; (5) enter the next frame signal judgment, until the end of communication.

7. The dynamic physical layer authentication system based on compensation mechanism according to claim 6, wherein, Based on the current time error term , the physical layer authentication model based on error feedback is used to dynamically update the kernel function bandwidth parameter of the current time The calculation formula is as follows: wherein, is the kernel bandwidth parameter at the next time instant, is the error term at the previous time instant, is the error term at the previous two time instants, is the proportional gain, is the integral gain, is the derivative gain, used to dynamically control the variation of the kernel bandwidth parameter, so that the system maintains the stability and response ability of the authentication model in a variable wireless environment.

8. The dynamic physical layer authentication system based on compensation mechanism according to claim 6, wherein, The form of the Gaussian kernel function is: 。 9. The dynamic physical layer authentication system based on compensation mechanism according to claim 6, wherein, In the trust phase, the initial Kalman filter state and the initial kernel function bandwidth parameter value are obtained, and the trust phase is trained by a known legal device for several rounds to build a basic model.

10. The dynamic physical layer authentication system based on compensation mechanism according to claim 6, wherein, with the authentication threshold The comparison formula is expressed as follows: "pass" indicates that the current authentication is passed, and the corresponding device is a legal device; "fail" indicates that the current authentication is failed, and the corresponding device is an illegal device.

Images