A privacy-protected intelligent security monitoring and protection method, device, and product

Abstract

The present invention discloses a privacy-protected intelligent security monitoring and protection method, device, and product. First, the transmitter transmits a WiFi data packet, and at the same time, the protector sends an obfuscated signal for obfuscation protection. The protector modifies its own MAC address to disguise itself as the transmitter and transmit a signal for dual-AP obfuscation protection. At the same time, an encryption algorithm is used to encrypt the CSI signal in the current space. Finally, the receiver captures the obfuscated WiFi data packet to sense the approach of people and provide early warning of their behavior. The present invention has a wide coverage range and economical deployment cost. By deploying dual wireless access points to transmit interference signals and combining a network authentication mechanism, it can effectively defend against CSI attacks, thereby improving security. In addition, the present invention uses a pre-trained human behavior recognition model to trace human behavior and provide audit support. At the same time, it does not rely on visual or sound data, avoids the use of stored data that may leak privacy, and further strengthens user privacy protection.

Description

Technical Field

[0001] The present invention belongs to the field of computer science and technology, and relates to an intelligent security monitoring and protection method, device, and product. Specifically, it relates to an intelligent security monitoring and protection method, device, and product that analyzes channel state information (CSI) in Wi-Fi signals to perceive human activities. Background Art

[0002] As a fundamental safety feature, security monitoring systems play a vital role in both work and life. Traditional security monitoring systems rely on wireless cameras, infrared sensors, and other devices for monitoring and alarms, playing an indispensable role in numerous locations, including homes, businesses, banks, and governments.

[0003] As technology advances, criminals' methods of committing crimes are constantly evolving. Traditional security monitoring systems primarily rely on wireless cameras to capture video data for tracing malicious behavior. However, this often involves sensitive personal information, posing the risk of theft and serious privacy breaches. Furthermore, to conserve storage space, systems may reduce video quality or periodically delete data, compromising tracing accuracy and data integrity. Furthermore, while using infrared sensors for monitoring can enhance security, limited detection range and external interference increase operating costs and reduce monitoring effectiveness. Furthermore, the increasing adoption of wireless technologies such as Wi-Fi in security monitoring systems, while streamlining many processes, also opens up the possibility for criminals to use CSI eavesdropping techniques to analyze indoor activity, creating new security challenges. Summary of the Invention

[0004] In view of the limitations of the above traditional security monitoring systems, the present invention provides an intelligent security monitoring and protection method, equipment and product with wide coverage and low deployment cost and privacy protection.

[0005] The technical solution adopted by the method of the present invention is: a privacy-protected intelligent security monitoring and protection method, comprising the following steps:

[0006] Step 1: The transmitter transmits a WiFi data packet, and the protector sends an obfuscation signal to perform obfuscation protection;

[0007] The protector modifies its own MAC address to disguise itself as a transmitter and transmit signals to prevent dual AP confusion. At the same time, it uses an encryption algorithm to encrypt the CSI signal in the current space.

[0008] Step 2: The receiver captures the obfuscated WiFi data packets and decodes them using the encryption matrix Ψ. It then extracts the CSI information from the data packets to detect human approach and provide early warning of human behavior. Finally, the analyzed data is uploaded to the client log.

[0009] Preferably, in step 1, the encryption formula of the encryption algorithm is:

[0010] y=HΨs+n;

[0011] Where y represents the received signal, H is the channel matrix, Ψ is the encryption matrix, s is the transmitted signal, and n is the noise.

[0012] Preferably, in step 2, the CSI trigger threshold and the time threshold are used to respectively determine whether a person is approaching and whether the approaching person stays for too long. If the two thresholds are exceeded, it can be determined that the current person has malicious behavior;

[0013] The CSI trigger threshold is calculated as follows:

[0014] σ threshold =λ·MAD(Z)+median(Z);

[0015] Where Z is a long-term record of the CSI signal amplitude, MAD(Z) refers to the mean absolute deviation of Z, λ is a parameter that controls the influence of the MAD value of the previous results on the current threshold, and median(Z) represents the median of Z.

[0016] When the CSI signal amplitude exceeds the CSI trigger threshold within a certain period of time, it indicates that someone is approaching, and the receiver will be triggered to record this period of time. The time threshold t_threshold is updated as follows:

[0017] t_threshold=p·t_threshold+q·kT;

[0018] Where T is the standard time length of a set of CSI signals, kT represents the duration of the malicious behavior, and p and q are parameters. If the duration of a CSI signal amplitude exceeds the threshold t_threshold, it means that the current person has stayed in the monitoring area for too long.

[0019] Preferably, in step 2, a human behavior recognition model is used to perform malicious behavior analysis;

[0020] The human behavior recognition model consists of a CSI signal encoder Es, a CSI decoder Ds and a linear analysis layer Linear connected in sequence;

[0021] The CSI signal encoder Es consists of a sequentially connected LSTM layer, a ReLU layer, a convolutional neural network (CNN) layer, and a self-attention mechanism (Self-attention) layer;

[0022] The CSI decoder Ds is composed of a structure consisting of four sequentially connected transposed convolutional layers (Trans Conv) + batch normalization layers (BatchNorm) + LeakyReLU. Finally, the data is further compressed and organized through a data transformation (Resize) layer for use in the next layer.

[0023] The linear analysis layer Linear is composed of a linear layer and a ReLU layer connected sequentially.

[0024] Preferably, the training process of the human action recognition model is divided into two steps. The first step is to use knowledge distillation to train Es and Ds using video information and CSI information. The second step is to add a linear layer after Es and Ds are trained, fix the Es and Ds parameters, and train the linear layer using CSI information and action information.

[0025] As a preference, in step 2, a human behavior recognition model is used to perform behavior analysis to obtain a specific behavior category INDEX. I , and record the first three actions that appear. Specifically, the following calculation method is used:

[0026] INDEX I =argmax(P I );

[0027] Where, P I is the action probability matrix, which records the probability of each action under each standard time series I. The argmax function is used to select the index INDEX of the behavior with the highest probability under the standard time series I. I The index corresponds to a specific behavior, and the corresponding action can be found in the specified action table according to the index.

[0028] The technical solution adopted by the device of the present invention is: a privacy-protected intelligent security monitoring and protection device, including a memory, a processor, and a computer program stored on the memory and runnable on the processor, and the processor implements the privacy-protected intelligent security monitoring and protection method when executing the program.

[0029] The technical solution adopted by the product of the present invention is: a computer program product, including a computer program, which implements the intelligent security monitoring and protection method for privacy protection when executed by a processor.

[0030] Compared with the prior art, the beneficial effects of the present invention include:

[0031] 1) User-awareness: This device can be easily installed and migrated, and during daily use, users do not need to perform additional operations or be aware of its presence. Users can continue to use the WiFi network normally without being affected or interfered with by the device. Its concealed nature means that users do not experience any additional burden or discomfort during use.

[0032] 2) Vision-Independent: This invention utilizes WiFi signals for monitoring and analysis, unaffected by light, obstructions, or other visual factors. Its functionality and effectiveness are independent of visual information, meaning the system can operate effectively in a variety of environments, with or without visual information.

[0033] 3) Privacy Protection: This invention places special emphasis on protecting user privacy, employing a variety of technical measures to ensure that user data and behavior information cannot be stolen or misused. For example, it uses anti-CSI eavesdropping analysis technology to protect privacy in indoor environments, and incorporates a behavior tracing and auditing module to monitor and record system operations, ensuring the effectiveness of privacy protection mechanisms. Users can use the device with confidence, without worrying about the leakage or misuse of their personal information. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] The technical solution of the present invention is further illustrated below using embodiments and specific implementation methods. In addition, some drawings are also used in the process of illustrating the technical solution. For those skilled in the art, other drawings and the intention of the present invention can be obtained based on these drawings without making any creative efforts.

[0035] Figure 1 is a schematic diagram of a method flow in an embodiment of the present invention;

[0036] Figure 2 4 is a structural diagram of a human behavior recognition model according to an embodiment of the present invention. DETAILED DESCRIPTION

[0037] In order to facilitate ordinary technicians in this field to understand and implement the present invention, the present invention is further described in detail below with reference to the accompanying drawings and examples. It should be understood that the implementation examples described herein are only used to illustrate and explain the present invention and are not used to limit the present invention.

[0038] Please see Figure 1 This embodiment provides a privacy-preserving intelligent security monitoring and protection method, comprising the following steps:

[0039] Step 1: The transmitter transmits a WiFi data packet, and the protector sends an obfuscation signal to perform obfuscation protection;

[0040] The protector modifies its own MAC address to disguise itself as a transmitter and transmit signals to prevent dual AP confusion. At the same time, it uses an encryption algorithm to encrypt the CSI signal in the current space.

[0041] In one embodiment, a specific encryption matrix Ψ is used to directly act on the transmitted signal, changing its characteristics to encrypt the data. Only authenticated devices can obtain this encryption matrix Ψ to decrypt its CSI information. The encryption algorithm is formulated as follows:

[0042] y=HΨs+n;

[0043] Where y represents the received signal, H is the channel matrix, Ψ is the encryption matrix, s is the transmitted signal, and n is the noise.

[0044] Step 2: The receiver captures the obfuscated WiFi data packets and decodes them using the encryption matrix Ψ. It then extracts the CSI information from the data packets to detect human approach and provide early warning of human behavior. Finally, the analyzed data is uploaded to the client log.

[0045] In one embodiment, a CSI trigger threshold and a time threshold are used to respectively determine whether a person is approaching and whether the approaching person stays for too long. If the two thresholds are exceeded, it can be determined that the current person has malicious behavior.

[0046] The CSI trigger threshold is calculated as follows:

[0047] σ threshold =λ·MAD(Z)+median(Z);

[0048] Where Z is a long-term record of the CSI signal amplitude, MAD(Z) refers to the mean absolute deviation of Z, λ is a parameter that controls the influence of the MAD value of the previous results on the current threshold, and median(Z) represents the median of Z.

[0049] When the CSI signal amplitude exceeds the CSI trigger threshold within a certain period of time, it indicates that someone is approaching, and the receiver will be triggered to record this period of time. The time threshold t_threshold is updated as follows:

[0050] t_threshold=p·t_threshold+q·kT;

[0051] Where T is the standard time length of a set of CSI signals, kT represents the duration of the malicious behavior, and p and q are parameters. If the duration of a CSI signal amplitude exceeds the threshold t_threshold, it means that the current person has stayed in the monitoring area for too long.

[0052] In one embodiment, a human behavior recognition model is used to perform malicious behavior analysis;

[0053] Please see Figure 2 The human behavior recognition model is composed of a CSI signal encoder Es, a CSI decoder Ds and a linear analysis layer Linear connected in sequence;

[0054] The CSI signal encoder Es consists of a sequentially connected LSTM layer, a ReLU layer, a convolutional neural network (CNN) layer, and a self-attention mechanism (Self-attention) layer;

[0055] The CSI decoder Ds is composed of a structure consisting of four sequentially connected transposed convolutional layers (Trans Conv) + batch normalization layers (BatchNorm) + LeakyReLU. Finally, the data is further compressed and organized through the data transformation (Resize) layer for use in the next layer.

[0056] The linear analysis layer Linear is composed of a linear layer and a ReLU layer connected sequentially.

[0057] In one embodiment, the training process of the human action recognition model is divided into two steps. The first step is to use knowledge distillation to train Es and Ds using video information and CSI information. The second step is to add a linear layer after Es and Ds are trained, fix the Es and Ds parameters, and train the linear layer using CSI information and action information.

[0058] In one embodiment, the occurrence time, end time and specific behavior of the malicious behavior are recorded in the log and sent to the client. The human behavior recognition model is used to perform behavior analysis to obtain the specific behavior category INDEX. I , and record the first three actions that appear. Specifically, the following calculation method is used:

[0059] INDEX I =argmax(P I );

[0060] Where, P I is the action probability matrix, which records the probability of each action under each standard time series I. The argmax function is used to select the index INDEX of the behavior with the highest probability under the standard time series I. I The index corresponds to a specific behavior, and the corresponding action can be found in the specified action table according to the index.

[0061] This embodiment also provides a privacy-protected intelligent security monitoring and protection device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, the privacy-protected intelligent security monitoring and protection method is implemented.

[0062] This embodiment also provides a computer program product, including a computer program, which implements the privacy protection intelligent security monitoring and protection method when executed by a processor.

[0063] The human action recognition model designed by the present invention combines knowledge distillation, autoencoders, and self-attention mechanisms during training. During training, the model is divided into a video encoder-decoder and a CSI signal encoder-decoder. The video encoder-decoder is a CNN-based video autoencoder that can downsample 28 human skeletal key points into a low-dimensional vector. This vector is used to supervise the downsampling process of the CSI signal encoder (Es) to ensure that the low-dimensional vector output by Es is as consistent as possible with the output of the video encoder (Ev). The vector output by Es can be effectively converted back into 28 human key points by the video decoder (Dv). The CSI signal encoder-decoder reuses the aforementioned video decoder (Dv) as its CSI decoder (Ds) and designs the CSI encoder (Es) as a combination of LSTM and CNN for conversion. In addition, to improve the recognition ability and differentiated processing of data features during the encoding process, both encoders (Es and Ev) integrate a self-attention mechanism. Finally, a linear layer is added after the CSI signal encoder-decoder Ds to perform action recognition.

[0064] In practical applications, the human behavior recognition model only requires the CSI signal encoder-decoder Ds and a linear layer to function properly. This training process is completed at the factory, eliminating the need for users to perform any training or configuration upon initial unboxing. Furthermore, manufacturers can continuously optimize and upgrade the model through remote updates.

[0065] The present invention proposes a privacy-protected intelligent security monitoring and protection method, device, and product that can use CSI information in WiFi signals to perceive human activities and provide high-precision warnings of approaching people, while also having the advantages of wide coverage and economical deployment. At the same time, in order to defend against CSI eavesdropping in indoor scenarios, the present invention utilizes indoor situational privacy protection technology that resists CSI eavesdropping analysis. By deploying fake AP devices to transmit interference signals and combining them with network authentication mechanisms, it can effectively defend against CSI attacks and improve system security. In addition, the present invention also uses privacy-protected behavior tracing and auditing technology, which can use pre-trained human behavior recognition models to trace human behavior and provide audit support. It does not rely on visual or sound data, avoids the use of stored data that can leak privacy, and further protects user privacy. These measures not only improve the security of the system, but also enhance users' trust in and acceptance of the technology.

[0066] It should be understood that the embodiments described above are only some of the embodiments of the present invention, rather than all of the embodiments. In addition, the technical features of the various embodiments or individual embodiments provided by the present invention may be arbitrarily combined with each other to form a feasible technical solution. Such combination is not restricted by the order of steps and / or structural composition mode, but must be based on the ability of ordinary technicians in this field to implement it. When the combination of technical solutions is mutually inconsistent or cannot be implemented, it should be deemed that such combination of technical solutions does not exist and is not within the scope of protection claimed by the present invention.

[0067] It should be understood that the above description of the preferred embodiment is relatively detailed and cannot be regarded as limiting the scope of protection of the patent of the present invention. Under the guidance of the present invention, ordinary technicians in this field can also make substitutions or modifications without departing from the scope of protection of the claims of the present invention, which all fall within the scope of protection of the present invention. The scope of protection requested by the present invention shall be based on the attached claims.

Claims

1. A privacy-preserving intelligent security monitoring and protection method, characterized in that: The following steps are involved: Step 1: The transmitter transmits a WiFi data packet, and the protector sends an obfuscation signal to perform obfuscation protection; The protector modifies its own MAC address to disguise itself as a transmitter and transmit signals to perform dual AP confusion protection; at the same time, it uses an encryption algorithm to encrypt the CSI signal in the current space; Step 2: The receiver captures the obfuscated WiFi data packets and decodes them using the encryption matrix Ψ. It then extracts the CSI information from the data packets to detect human approach and provide early warning of human behavior. Among them, the CSI trigger threshold and time threshold are used to determine whether there is a person approaching and whether the approaching person stays for too long. If the above two thresholds are exceeded, it is determined that the current person has malicious behavior.

2. The privacy protection intelligent security monitoring and protection method according to claim 1 is characterized by: In step 1, the encryption formula of the encryption algorithm is: y=HΨs+n; Where y represents the received signal, H is the channel matrix, Ψ is the encryption matrix, s is the transmitted signal, and n is the noise.

3. The privacy protection intelligent security monitoring and protection method according to claim 1 is characterized by: In step 2, the CSI trigger threshold is calculated as follows: σ threshold =λ·MAD(Z)+median(Z); Where Z is a long-term record of the CSI signal amplitude, MAD(Z) refers to the mean absolute deviation of the sequence Z; λ is a parameter that controls the influence of the MAD value of the previous results on the current threshold; median(Z) represents the median of the sequence Z. When the CSI signal amplitude exceeds the CSI trigger threshold within a certain period of time, it indicates that someone is approaching, and the receiver will be triggered to record this period of time. The time threshold t_threshold is updated as follows: t_threshold=p·t_threshold+q·kT; Where T is the standard time length of a set of CSI signals, kT represents the duration of the malicious behavior, and p and q are parameters. If the duration of the CSI signal amplitude exceeds the threshold t_threshold, it means that the current person has stayed in the monitoring area for too long.

4. The privacy protection intelligent security monitoring and protection method according to claim 1 is characterized by: In step 2, a human behavior recognition model is used to analyze malicious behavior; The human behavior recognition model consists of a CSI signal encoder Es, a CSI decoder Ds and a linear analysis layer Linear connected in sequence; The CSI signal encoder Es is composed of a sequentially connected LSTM layer, a ReLU layer, a convolutional neural network (CNN) layer, and a self-attention mechanism layer. The CSI decoder Ds is composed of a structure consisting of four sequentially connected transposed convolutional layers (Trans Conv), batch normalization layers (BatchNorm), and LeakyReLU. Finally, the data is further compressed and organized through the data transformation layer (Resize) for use in the next layer. The linear analysis layer Linear is composed of a linear layer and a ReLU layer connected sequentially.

5. The privacy protection intelligent security monitoring and protection method according to claim 4 is characterized by: The training process of the human behavior recognition model is divided into two steps. The first step is to use the knowledge distillation method to train Es and Ds using video information and CSI information. The second step is to add a Linear layer to the model after Es and Ds are trained, fix the Es and Ds parameters, and train the Linear layer using CSI information and motion information.

6. The privacy protection intelligent security monitoring and protection method according to any one of claims 1 to 5, characterized in that: In step 2, the human behavior recognition model is used to perform behavior analysis to obtain the specific behavior category INDEX I , and record the first three actions that appear; specifically, use the following calculation method: INDEX I =argmax(P I ); Where, P I is the action probability matrix, which records the probability of each action under each standard time series I. The argmax function is used to select the index INDEX of the behavior with the highest probability under the standard time series I. I , the index corresponds to a specific behavior, and according to the index, the corresponding action is found in the specified action table.

7. A privacy-protected intelligent security monitoring and protection device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the intelligent security monitoring and protection method for privacy protection as described in any one of claims 1 to 6 is implemented.

8. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the intelligent security monitoring and protection method for privacy protection as claimed in any one of claims 1 to 6 is implemented.

Images