Privacy protection wireless sensing method and system based on CSI (Channel State Information) confusion

By introducing CSI obfuscation technology into Wi-Fi signals, the key sharing and response prediction network are used to restore the original CSI, solving the problem of privacy protection in wireless perception of Wi-Fi signals, achieving efficient privacy protection and legal perception.

CN118921656APending Publication Date: 2024-11-08SOUTHEAST UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410970051.0
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-19
Publication Date
2024-11-08

AI Technical Summary

Technical Problem

The original CSI of Wi-Fi signals is directly used in wireless perception and is susceptible to third-party attacks, and cannot take into account both legal wireless perception and privacy protection.

Method used

Using a CSI obfuscation method, the obfuscation pilot signal is generated through key sharing technology, and the original CSI is restored using time-varying filters and response prediction networks to ensure the perceived ability of the legal receiver and at the same time reduce the perceived performance of the eavesdropper.

Benefits of technology

It realizes that while maintaining more than 90% of the legal sensing performance, the sensor performance of the eavesdropper is reduced to less than 50%, ensuring privacy protection.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118921656A_ABST
    Figure CN118921656A_ABST
Patent Text Reader

Abstract

The invention discloses a privacy protection wireless sensing method and system based on CSI confusion, and the method comprises the steps: (1), a transmitting end and a receiving end respectively obtain a secret key kt through a secret key sharing technology; (2) the transmitting end takes the key kt as a confusion coefficient of a confusion device, confuses the pilot signal through the confusion device to generate a confused pilot signal, and sends the confused pilot signal to a receiving end in a WIFI environment where a sensing target is located; (3) the receiving end carries out channel estimation according to the received confusion pilot signals to obtain confusion CSI; (4) the receiving end calculates a confusion coefficient vector F = f (ktHkt) according to the key kt; (5) the receiving end inputs the confusion coefficient vector F into the trained response prediction network to obtain a confusion response estimated value # imgabs0 # of the confusion device; (6) the receiving end divides the confusion CSI by the confusion response estimated value # imgabs1 # to obtain original CSI without the confusion device; and (7) the receiving end performs wireless sensing of the sensing target according to the original CSI. According to the method, legal wireless sensing and privacy protection are considered at the same time.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to wireless sensing technology, and in particular to a privacy protection wireless sensing method and system based on CSI obfuscation. Background Art

[0002] Wireless sensing is a sensing technology based on wireless signals. It mainly senses the behavior of the human body, such as position and posture, by analyzing the changing characteristics of wireless signals when passing through the human body. People do not need to wear any equipment and can achieve non-invasive detection without being aware of it. Wi-Fi is a wireless local area network technology created based on the IEEE 802.11 standard. Wireless network in the scope of wireless local area network refers to "wireless compatibility certification". In the past, network cables were responsible for networking, while Wi-Fi uses radio waves to connect to the network. Within the effective range of the radio wave coverage of this wireless router, radio waves can be used to connect to the network. Therefore, with the rapid development and maturity of Wi-Fi technology, using Wi-Fi signals for wireless sensing not only does not require the deployment of additional equipment, but wireless sensing can also be carried out simultaneously with network communication, improving the utilization of wireless resources.

[0003] The human behavior perception technology based on wireless signals utilizes the influence of attenuation, reflection, diffraction and multipath effects in the propagation process of wireless signals. The specific behavior of the human body will cause specific regular interference to the wireless signal path. Therefore, this technology mainly completes behavior recognition by analyzing the regular changes of the signal. Channel State Information (CSI) describes the attenuation factor of the signal on each transmission path, that is, the value of each element in the channel gain matrix H, such as signal scattering, environmental attenuation, distance attenuation and other information, which can show the fluctuation trend of the signal under different interference levels, thereby expanding the perception range, improving the perception sensitivity, and increasing the robustness of perception. Therefore, many researchers and scholars have performed wireless perception based on the CSI of Wi-Fi signals.

[0004] However, Wi-Fi is open. If the original CSI of the Wi-Fi signal is used for direct wireless sensing, it is easy to be attacked by a third party. After the third party eavesdrops on the CSI, it can parse the target's private data. It is impossible to take into account both legal wireless sensing and privacy protection at the same time. Summary of the invention

[0005] In view of the problems existing in the prior art, the purpose of the present invention is to provide a privacy-preserving wireless sensing method and system based on CSI obfuscation that can take into account both legitimate wireless sensing and privacy protection.

[0006] In order to achieve the above-mentioned object of the invention, the present invention provides the following technical solutions:

[0007] A privacy protection wireless sensing method based on CSI obfuscation includes the following steps:

[0008] (1) The transmitter and receiver obtain the key k through key sharing technology. t =[1,b 1t ,...,b Nt ] T , b 1t ,...,b Nt Denote the key k t The 2nd,…,Nth coefficients in , where N is a positive integer greater than or equal to 2;

[0009] (2) The transmitter sends the key k t As the confusion coefficient of the obfuscator, the pilot signal is obfuscated by the obfuscator to generate an obfuscated pilot signal, and is sent to the receiving end in the WIFI environment where the sensing target is located;

[0010] (3) The receiving end performs channel estimation based on the received confused pilot signal to obtain confused CSI;

[0011] (4) The receiving end uses the key k t The confusion coefficient vector F = f(k t H k t ), where f() represents a mapping function that converts a matrix into a vector;

[0012] (5) The receiver inputs the confusion coefficient vector F into the trained response prediction network to obtain the confusion response estimate of the obfuscator. Wherein, the response prediction network is used to predict the estimated value of the obfuscator's obfuscator response according to the obfuscation coefficient vector;

[0013] (6) The receiver divides the obfuscated CSI by the obfuscated response estimate Get the original CSI without the obfuscator;

[0014] (7) The receiving end performs wireless perception of the target based on the original CSI.

[0015] Furthermore, the obfuscator is specifically a time-varying filter FIR.

[0016] Furthermore, the response prediction network is specifically RPNet.

[0017] Furthermore, the response prediction network is trained using a number of sample pairs, and the sample pairs are obtained through the following steps:

[0018] A. The transmitter and receiver obtain several different test keys k through key sharing technology.t,i , i=1,2,…,M, M is the number of test keys;

[0019] B. In a secure WIFI environment, the transmitter turns off the obfuscator and transmits the pilot signal directly to the receiver;

[0020] C. The receiving end obtains the original CSI C0 through signal estimation based on the received pilot signal;

[0021] D. The transmitter sends each test key k in turn. t,i As the confusion coefficient of the obfuscator, the pilot signal is obfuscated by the obfuscators with different confusion coefficients to generate the test obfuscated pilot signal D i , i = 1, 2, ..., M, sent to the receiving end in a secure WIFI environment;

[0022] E. The receiving end confuses the pilot signal D according to the received test i Perform channel estimation and obtain the corresponding test confusion CSI CX i , i=1,2,…,M;

[0023] F. The receiving end uses each test key k t,i Calculate the corresponding test confusion coefficient vector F i =f(k t, i H k t,i ),i=1,2,…,M;

[0024] G. Confusing the test with CSI CX i Divide by the original CSI C0 to get the corresponding test confusion response estimate

[0025] H, each test confusion coefficient vector F i As a sample, the corresponding test confounded response estimate As labels, generate several sample pairs.

[0026] Furthermore, in step (7), the receiving end uses a neural network model to perform wireless perception of the perception target according to the original CSI.

[0027] A privacy protection wireless sensing system based on CSI obfuscation includes a transmitting end and a receiving end, wherein:

[0028] The transmitting end comprises:

[0029] The key sharing module is used to obtain the key k through the key sharing technology with the receiving end. t =[1,b 1t ,...,bNt ] T , b 1t ,...,b Nt Denote the key k t The 2nd,…,Nth coefficients in , where N is a positive integer greater than or equal to 2;

[0030] Obfuscator, used to convert the key k t As the confusion coefficient, the pilot signal is confused to generate a confused pilot signal;

[0031] A signal sending module is used to send a confusing pilot signal to a receiving end in a WIFI environment where a sensing target is located;

[0032] The receiving end comprises:

[0033] The key sharing module is used to obtain the key k through the key sharing technology with the transmitter. t =[1,b 1t ,...,b Nt ] T ;

[0034] A channel estimation module, used to perform channel estimation based on the received confused pilot signal to obtain confused CSI;

[0035] The obfuscation coefficient vector calculation module is used to calculate the obfuscation coefficient vector according to the key k t The confusion coefficient vector F = f(k t H k t ), where f() represents a mapping function that converts a matrix into a vector;

[0036] The response prediction module is used to input the confusion coefficient vector F into the trained response prediction network to obtain the confusion response estimate of the obfuscator. Wherein, the response prediction network is used to predict the estimated value of the obfuscator's obfuscator response according to the obfuscation coefficient vector;

[0037] The original CSI estimation module is used to divide the obfuscated CSI by the obfuscated response estimate Get the original CSI without the obfuscator;

[0038] The wireless sensing module is used to perform wireless sensing of the sensing target according to the original CSI.

[0039] Furthermore, the obfuscator is specifically a time-varying filter FIR.

[0040] Furthermore, the response prediction network is specifically RPNet.

[0041] Furthermore, the response prediction network is trained using a number of sample pairs, and the sample pairs are obtained through the following steps:

[0042] A. The transmitter and receiver obtain several different test keys k through key sharing technology. t,i , i=1,2,…,M, M is the number of test keys;

[0043] B. In a secure WIFI environment, the transmitter turns off the obfuscator and transmits the pilot signal directly to the receiver;

[0044] C. The receiving end obtains the original CSI C0 through signal estimation based on the received pilot signal;

[0045] D. The transmitter sends each test key k in turn. t,i As the confusion coefficient of the obfuscator, the pilot signal is obfuscated by the obfuscators with different confusion coefficients to generate the test obfuscated pilot signal D i , i = 1, 2, ..., M, sent to the receiving end in a secure WIFI environment;

[0046] E. The receiving end confuses the pilot signal D according to the received test i Perform channel estimation and obtain the corresponding test confusion CSI CX i , i=1,2,…,M;

[0047] F. The receiving end uses each test key k t,i Calculate the corresponding test confusion coefficient vector F i =f(k t, i H k t,i ),i=1,2,…,M;

[0048] G. Confusing the test with CSI CX i Divide by the original CSI C0 to get the corresponding test confusion response estimate

[0049] H, each test confusion coefficient vector F i As a sample, the corresponding test confounded response estimate As labels, generate several sample pairs.

[0050] Furthermore, the wireless sensing module uses a neural network model to perform wireless sensing of the sensing target according to the original CSI.

[0051] Compared with the prior art, the present invention has the following beneficial effects: the present invention proposes a privacy-preserving wireless sensing method and system based on CSI obfuscation, adopts dynamic channel obfuscation technology at the transmitting end, and restores the sensing capability at the legitimate receiving end. Compared with the prior art, the present invention takes into account both legitimate sensing and can prevent illegal attacks to achieve privacy protection. Simulation results show that the present invention can reduce the sensing performance of the eavesdropper to less than 50%, while maintaining the legitimate sensing performance above 90%. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 It is a flowchart of a privacy protection wireless sensing method based on CSI obfuscation provided by an embodiment of the present invention;

[0053] Figure 2 It is a schematic diagram of an experimental environment arrangement of the present invention;

[0054] Figure 3 It is a schematic diagram of a process of providing gains for different subcarriers to blur original response values ​​according to an embodiment of the present invention;

[0055] Figure 4 It is the hardware offset problem that makes demixing difficult at the legitimate receiving end in the embodiment of the present invention;

[0056] Figure 5 This is the privacy protection perception process provided by the embodiment of the present invention;

[0057] Figure 6 It is a schematic diagram of the recovery effect of the method in an embodiment of the present invention on actual data. DETAILED DESCRIPTION

[0058] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0059] Embodiment 1

[0060] The embodiment of the present invention provides a privacy protection wireless sensing method based on CSI obfuscation, such as Figure 1 As shown, the following steps are included:

[0061] (1) The transmitter and receiver obtain the key k through key sharing technology. t .

[0062] Assume that the transmitter is Alice, the legitimate receiver is Bob, the eavesdropping object is the sensing target Vic, and the eavesdropper who performs illegal sensing activities is Eve, such as Figure 2 As shown, both the transmitter and the receiver are software radio devices capable of running openWi-Fi.

[0063] In the specific implementation, the construction form of the key sharing technology is not subject to specific restrictions. It only needs to ensure that the transmitter and the receiver have the same key k. t The specific construction methods include preset sequence, Diffie-Hellman protocol, KDC system and IBKE method. t =[1,b 1t ,...,b Nt ] T , b 1t ,...,b Nt Denote the key k t The 2nd,…,Nth coefficients in , where N is a positive integer greater than or equal to 2.

[0064] (2) The transmitter sends the key k t As the confusion coefficient of the obfuscator, the pilot signal is obfuscated by the obfuscator to generate a confused pilot signal, and is sent to the receiving end in the WIFI environment where the sensing target is located.

[0065] The obfuscator is used to provide different gains for different subcarriers to blur the original response value, so that the CSI is distorted, such as multiplying by different factors. Therefore, there are many specific implementation methods of the obfuscator, such as distorting the I / Q signal by a time-varying filter FIR built into its transmitter. t As the confusion coefficient of the confuser, the tap coefficients of the time-varying filter FIR are set to 1, b 1t ,...,b Nt ,like Figure 3 Before the pilot signal is sent, it first passes through a time-varying filter FIR, that is, the pilot signal is multiplied by and then sent to the receiving end.

[0066] (3) The receiving end performs channel estimation based on the received confused pilot signal to obtain the confused CSI.

[0067] At this time, for the eavesdropper Eve: by applying 1,b 1t ,...,b Nt Multiple confusion coefficients change the gain on different subcarriers, which greatly changes the CSI. Even if the confused pilot signal is eavesdropped and the confused CSI is estimated, it is still difficult to obtain the correct CSI because there is no key k. t , it is impossible to de-mix the obfuscated CSI and extract the original CSI, and it is also impossible to obtain the target’s private data, thereby achieving the purpose of privacy protection.

[0068] (4) The receiving end uses the key k t The confusion coefficient vector F = f(k t H k t ), where f() represents the mapping function that converts the matrix into a vector.

[0069] For the legitimate receiver: To recover the original CSI from the obfuscated CSI, it is necessary to find the 1t ,...,b Nt ]The corresponding obfuscator working mode is the obfuscation response G.

[0070] Since G can be approximately linearly expressed by the confusion coefficient:

[0071]

[0072] W is the linear transformation matrix, S is the number of carriers, and w(i) = [u 11 ,…,u (N+1)1 ] represents the i-th row vector of W (i∈1, 2, ..., S).

[0073] However, due to hardware offsets, see Figure 4 As shown, it makes it difficult for the legitimate receiving end to demix. However, the real and imaginary parts still maintain a linear relationship with the confusion coefficient, so in order to obtain an estimate of the confusion response G Consider mapping the confusion coefficient to multiple features as a measure of amplitude. Taking the modulus of G, that is, taking the modulus of the real and imaginary parts of the linear relationship, the amplitude of the aliasing response on the i-th subcarrier at time t can be expressed as

[0074] ‖G(i)‖=(w(i) H k t H k t w(i)) 1 / 2

[0075] For example, when N = 2, k t =[1,b 1t ,b 2t ] T ,but

[0076]

[0077] From the above formula, we can see that we only need to fit w(i) and k t H k t This nonlinear relationship can restore the required amplitude. t H k t Mapped into a vector combination, as input, to obtain through nonlinear relationship For example, when N = 2, k t H k t Mapped Vector

[0078] (5) The receiver inputs the confusion coefficient vector F into the trained response prediction network to obtain the confusion response estimate of the obfuscator. The response prediction network is used to predict the estimated value of the obfuscator's obfuscator response according to the obfuscation coefficient vector.

[0079] Here, the present invention uses a deep neural network to capture the relationship between subcarriers and confusion coefficients, and designs a response prediction network RPNet, such as Figure 5 As shown, the estimated value of the confusion response can be accurately predicted. The densely connected layer in RPNet acts as a feature extractor. There are many ways to build RPNet, such as using a GRU layer with a hidden state size of 256 and a densely connected layer containing 512, 256, 128 and 64 neurons respectively, or a LSTM network and RNN network with similar principles. For example, when N = 2, the input confusion coefficient vector

[0080] The response prediction network is trained using a number of sample pairs, and the sample pairs are obtained through the following steps:

[0081] A. The transmitter and receiver obtain several different test keys k through key sharing technology. t,i , i=1,2,…,M, M is the number of test keys;

[0082] B. In a secure WIFI environment, the transmitter turns off the obfuscator and transmits the pilot signal directly to the receiver;

[0083] C. The receiving end obtains the original CSI C0 through signal estimation based on the received pilot signal;

[0084] D. The transmitter sends each test key k in turn. t,i As the confusion coefficient of the obfuscator, the pilot signal is obfuscated by the obfuscators with different confusion coefficients to generate the test obfuscated pilot signal D i , i = 1, 2, ..., M, sent to the receiving end in a secure WIFI environment;

[0085] E. The receiving end confuses the pilot signal D according to the received test i Perform channel estimation and obtain the corresponding test confusion CSI CX i , i=1,2,…,M;

[0086] F. The receiving end uses each test key k t,i Calculate the corresponding test confusion coefficient vector F i =f(k t, i H kt,i ),i=1,2,…,M;

[0087] G. Confusing the test with CSI CX i Divide by the original CSI C0 to get the corresponding test confusion response estimate

[0088] H, each test confusion coefficient vector F i As a sample, the corresponding test confounded response estimate As labels, generate several sample pairs.

[0089] (6) The receiving end will confuse CSI F a Divide by the confounded response estimate Get the original CSIM without the obfuscator a .

[0090] (7) The receiving end performs wireless perception of the target based on the original CSI.

[0091] The receiving end uses a neural network model to perform wireless perception of the perception target based on the original CSI, such as a deep neural network, a convolutional neural network, RPNet, etc., to obtain the perception classification result of the perception target. For example, when the perception purpose is to perceive the position, the different positions of the perception target are used as classification targets to obtain the classification result, and finally determine the positioning information of the perception target.

[0092] Finally, the present invention is verified by simulation.

[0093] First, to measure G and The similarity between , using cosine similarity as the performance indicator:

[0094]

[0095] Finally, the predicted response was found The cosine similarity with the actual response G is above 0.99, which means that this method is feasible.

[0096] Secondly, the packet error rate is used to evaluate the method. Figure 6 As shown in Figure 1, we collected positioning data from two receivers, before the obfuscator was activated and after the obfuscator was activated. Finally, when there was no obfuscation, the overall recognition rate was always over 90%. After obfuscation, the accuracy of the two receivers dropped by more than 50%, which shows that the method of the present invention can effectively prevent the attack of the eavesdropper Eve.

[0097] Embodiment 2

[0098] The embodiment of the present invention provides a privacy protection wireless sensing system based on CSI obfuscation, including a transmitting end and a receiving end, wherein:

[0099] The transmitting end comprises:

[0100] The key sharing module is used to obtain the key k through the key sharing technology with the receiving end. t =[1,b 1t ,...,b Nt ] T , b 1t ,...,b Nt Denote the key k t The 2nd,…,Nth coefficients in , where N is a positive integer greater than or equal to 2;

[0101] Obfuscator, used to convert the key k t As the confusion coefficient, the pilot signal is confused to generate a confused pilot signal; specifically, a time-varying filter FIR;

[0102] A signal sending module is used to send a confusing pilot signal to a receiving end in a WIFI environment where a sensing target is located;

[0103] The receiving end comprises:

[0104] The key sharing module is used to obtain the key k through the key sharing technology with the transmitter. t =[1,b 1t ,...,b Nt ] T ;

[0105] A channel estimation module, used to perform channel estimation based on the received confused pilot signal to obtain confused CSI;

[0106] The obfuscation coefficient vector calculation module is used to calculate the obfuscation coefficient vector according to the key k t The confusion coefficient vector F = f(k t H k t ), where f() represents a mapping function that converts a matrix into a vector;

[0107] The response prediction module is used to input the confusion coefficient vector F into the trained response prediction network to obtain the confusion response estimate of the obfuscator. Wherein, the response prediction network is used to predict the estimated value of the obfuscator's obfuscator response according to the obfuscation coefficient vector;

[0108] The original CSI estimation module is used to divide the obfuscated CSI by the obfuscated response estimate Get the original CSI without the obfuscator;

[0109] The wireless sensing module is used to sense the wireless of the sensing target according to the original CSI. The wireless sensing module uses a neural network model to sense the wireless of the sensing target according to the original CSI.

[0110] The obfuscator.

[0111] The response prediction network is specifically RPNet, which is trained using a number of sample pairs obtained by the following steps:

[0112] A. The transmitter and receiver obtain several different test keys k through key sharing technology. t,i , i=1,2,…,M, M is the number of test keys;

[0113] B. In a secure WIFI environment, the transmitter turns off the obfuscator and transmits the pilot signal directly to the receiver;

[0114] C. The receiving end obtains the original CSI C0 through signal estimation based on the received pilot signal;

[0115] D. The transmitter sends each test key k in turn. t,i As the confusion coefficient of the obfuscator, the pilot signal is obfuscated by the obfuscators with different confusion coefficients to generate the test obfuscated pilot signal D i , i = 1, 2, ..., M, sent to the receiving end in a secure WIFI environment;

[0116] E. The receiving end confuses the pilot signal D according to the received test i Perform channel estimation and obtain the corresponding test confusion CSI CX i , i=1,2,…,M;

[0117] F. The receiving end uses each test key k t,i Calculate the corresponding test confusion coefficient vector F i =f(k t, i H k t,i ),i=1,2,…,M;

[0118] G. Confusing the test with CSI CX i Divide by the original CSI C0 to get the corresponding test confusion response estimate

[0119] H, each test confusion coefficient vector F i As a sample, the corresponding test confounded response estimate As labels, generate several sample pairs.

[0120] The system provided in the embodiment of the present invention can be used to execute the method provided in the first embodiment of the present invention, and has the corresponding functions and beneficial effects of the execution method.

[0121] The system can be implemented in software and / or hardware, and the system can be configured in a terminal device.

[0122] It is worth noting that in the embodiment of the above system, the various units and modules included are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be achieved; in addition, the specific names of the functional units are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of the present invention.

[0123] It should be understood that the above embodiments and descriptions only describe the principles, main features and advantages of the present invention. Without departing from the spirit and scope of the present invention, the present invention may be subject to various changes and improvements, and these changes and improvements all fall within the scope of protection of the present invention.

Claims

1. A privacy-preserving wireless sensing method based on CSI obfuscation, characterized in that: The steps include: (1) The transmitter and receiver obtain the key k through key sharing technology. t =[1,b 1t ,...,b Nt ] T , b 1t ,...,b Nt Represent the key k t The 2nd,…,Nth coefficients in , where N is a positive integer greater than or equal to 2; (2) The transmitter sends the key k t As the confusion coefficient of the obfuscator, the pilot signal is obfuscated by the obfuscator to generate an obfuscated pilot signal, and is sent to the receiving end in the WIFI environment where the sensing target is located; (3) The receiving end performs channel estimation based on the received confused pilot signal to obtain confused CSI; (4) The receiving end uses the key k t The confusion coefficient vector F = f(k t H k t ), where f() represents a mapping function that converts a matrix into a vector; (5) The receiver inputs the confusion coefficient vector F into the trained response prediction network to obtain the confusion response estimate of the obfuscator. Wherein, the response prediction network is used to predict the estimated value of the obfuscator's obfuscator response according to the obfuscation coefficient vector; (6) The receiver divides the obfuscated CSI by the obfuscated response estimate Get the original CSI without the obfuscator; (7) The receiving end performs wireless perception of the target based on the original CSI.

2. The privacy protection wireless sensing method based on CSI obfuscation according to claim 1, characterized in that: The aliasing device is specifically a time-varying filter FIR.

3. The privacy protection wireless sensing method based on CSI obfuscation according to claim 1, characterized in that: The response prediction network is specifically RPNet.

4. The privacy protection wireless sensing method based on CSI obfuscation according to claim 1, characterized in that: The response prediction network is trained using a number of sample pairs, which are obtained through the following steps: A. The transmitter and receiver obtain several different test keys k through key sharing technology. t,i , i=1,2,…,M, M is the number of test keys; B. In a secure WIFI environment, the transmitter turns off the obfuscator and transmits the pilot signal directly to the receiver; C. The receiving end obtains the original CSIC0 through signal estimation based on the received pilot signal; D. The transmitter sends each test key k in turn. t,i As the confusion coefficient of the obfuscator, the pilot signal is obfuscated by the obfuscators with different confusion coefficients to generate the test obfuscated pilot signal D i , i = 1, 2, ..., M, sent to the receiving end in a secure WIFI environment; E. The receiving end confuses the pilot signal D according to the received test i Perform channel estimation and obtain the corresponding test confusion CSICX i , i=1,2,…,M; F. The receiving end uses each test key k t,i Calculate the corresponding test confusion coefficient vector F i =f(k t,i H k t,i ),i=1,2,…,M; G. Confusing the test with CSI CX i Divide by the original CSI C0 to get the corresponding test confusion response estimate i=1,2,…,M; H, each test confusion coefficient vector F i As a sample, the corresponding test confounded response estimate As labels, generate several sample pairs.

5. The privacy protection wireless sensing method based on CSI obfuscation according to claim 1, characterized in that: In step (7), the receiving end uses a neural network model to perform wireless perception of the perception target based on the original CSI.

6. A privacy-preserving wireless sensing system based on CSI obfuscation, characterized in that: It includes a transmitting end and a receiving end, wherein: The transmitting end comprises: The key sharing module is used to obtain the key k through the key sharing technology with the receiving end. t =[1,b 1t ,...,b Nt ] T , b 1t ,...,b Nt Represent the key k t The 2nd,…,Nth coefficients in , where N is a positive integer greater than or equal to 2; Obfuscator, used to convert the key k t As the confusion coefficient, the pilot signal is confused to generate a confused pilot signal; A signal sending module is used to send a confusing pilot signal to a receiving end in a WIFI environment where a sensing target is located; The receiving end comprises: The key sharing module is used to obtain the key k through the key sharing technology with the transmitter. t =[1,b 1t ,...,b Nt ] T ; A channel estimation module, used to perform channel estimation based on the received confused pilot signal to obtain confused CSI; The obfuscation coefficient vector calculation module is used to calculate the obfuscation coefficient vector according to the key k t The confusion coefficient vector F = f(k t H k t ), where f() represents a mapping function that converts a matrix into a vector; The response prediction module is used to input the confusion coefficient vector F into the trained response prediction network to obtain the confusion response estimate of the obfuscator. Wherein, the response prediction network is used to predict the estimated value of the obfuscator's obfuscator response according to the obfuscation coefficient vector; The original CSI estimation module is used to divide the obfuscated CSI by the obfuscated response estimate Get the original CSI without the obfuscator; The wireless sensing module is used to perform wireless sensing of the sensing target according to the original CSI.

7. The privacy protection wireless sensing system based on CSI obfuscation according to claim 6, characterized in that: The aliasing device is specifically a time-varying filter FIR.

8. The privacy protection wireless sensing system based on CSI obfuscation according to claim 6, characterized in that: The response prediction network is specifically RPNet.

9. The privacy protection wireless sensing system based on CSI obfuscation according to claim 6, characterized in that: The response prediction network is trained using a number of sample pairs, which are obtained through the following steps: A. The transmitter and receiver obtain several different test keys k through key sharing technology. t,i , i=1,2,…,M, M is the number of test keys; B. In a secure WIFI environment, the transmitter turns off the obfuscator and transmits the pilot signal directly to the receiver; C. The receiving end obtains the original CSIC0 through signal estimation based on the received pilot signal; D. The transmitter sends each test key k in turn. t,i As the confusion coefficient of the obfuscator, the pilot signal is obfuscated by the obfuscators with different confusion coefficients to generate the test obfuscated pilot signal D i , i = 1, 2, ..., M, sent to the receiving end in a secure WIFI environment; E. The receiving end confuses the pilot signal D according to the received test i Perform channel estimation and obtain the corresponding test confusion CSICX i , i=1,2,…,M; F. The receiving end uses each test key k t,i Calculate the corresponding test confusion coefficient vector F i =f(k t,i H k t,i ),i=1,2,…,M; G. Confusing the test with CSI CX i Divide by the original CSI C0 to get the corresponding test confusion response estimate i=1,2,…,M; H, each test confusion coefficient vector F i As a sample, the corresponding test confounded response estimate As labels, generate several sample pairs.

10. The privacy protection wireless sensing system based on CSI obfuscation according to claim 6, characterized in that: The wireless sensing module uses a neural network model to perform wireless sensing of the sensing target according to the original CSI.