A User Privacy Protection Method and Device for a Face Recognition System

Abstract

The present invention discloses a user privacy protection method and device for a face recognition system, belonging to the technical field of mobile device and software security. The method includes parallel processing of edge detection and lightweight semantic segmentation on an input image to generate an edge intensity map and a semantic confidence map; dynamically fusing the edge intensity map and the semantic confidence map through dynamic weights to generate a privacy classification mask; dividing the image into three levels of privacy regions through the privacy classification mask; and performing hierarchical encryption on the three levels of privacy regions through three pre-generated aperiodic chaotic sequences to complete user privacy protection. Among them, the three aperiodic chaotic sequences are iteratively generated through a Young's hyperchaotic system based on the hash value of the concatenation of a user-defined key and the system time. The present invention realizes image encryption according to privacy levels and protects user privacy in face recognition.

Description

Technical Field

[0001] The present invention relates to a user privacy protection method and device for a face recognition system, belonging to the technical field of mobile device and software security. Background Art

[0002] In recent years, with the development of large-scale modeling technology, the human face has become an important biometric feature, and facial recognition systems are widely used, especially on mobile devices. However, when interacting with the cloud, there is a risk of leaking user privacy, not only facial features but also background features.

[0003] Existing face protection methods are mainly based on image encryption or noise addition. They have the following problems:

[0004] 1. Existing image encryption relies on high-order hyperchaos and multiple rounds of iterations, resulting in low encryption efficiency and high computing power requirements, making it unsuitable for deployment on mobile devices and low-computing-power smart devices.

[0005] 2. The noise-addition protection method is irreversible and cannot be used in scenarios where decryption is required, such as data set collection and network video transmission.

[0006] 3. Security protection measures focus on preventing theft and forgery of faces, but do not protect the digital footprints and privacy information carried in the background area outside the face. Summary of the Invention

[0007] The purpose of the present invention is to overcome the deficiencies in the prior art and provide a user privacy protection method and device for a face recognition system, which realizes image encryption according to privacy levels and protects user privacy in face recognition.

[0008] To achieve the above object, the present invention is implemented by adopting the following technical solutions:

[0009] In a first aspect, the present invention provides a user privacy protection method for a face recognition system, comprising:

[0010] Perform edge detection and lightweight semantic segmentation on the input image in parallel to generate edge strength map and semantic confidence map;

[0011] The privacy graded mask is generated by fusing edge strength map and semantic confidence map through dynamic weights.

[0012] The image is divided into three levels of privacy areas through privacy classification masks;

[0013] Through the three pre-generated periodic chaotic sequences, the three-level privacy areas are hierarchically encrypted to achieve user privacy protection;

[0014] Among them, the three aperiodic chaotic sequences are generated by iterating through the Young's hyperchaotic system based on the hash value obtained by concatenating the user-defined key and the system time.

[0015] Further, the generation of the privacy grading mask by dynamically weighting and fusing the edge intensity map and the semantic confidence map includes:

[0016] Based on the local semantic confidence mean and the edge intensity standard deviation , calculate the fusion weight of semantic segmentation and edge detection. The formula is as follows:

[0017] ;

[0018] Among them, is the fusion weight of semantic segmentation, is the fusion weight of edge detection;

[0019] Generate a fusion mask through the weighting formula . Among them, is the semantic confidence map, is the edge intensity map.

[0020] Further, the division of the image into three levels of privacy regions by the privacy grading mask includes:

[0021] According to the threshold range into which the fusion mask falls, divide the image into three levels of privacy regions: the face feature area, the face contour area, and the background area.

[0022] Further, the generation method of the three aperiodic chaotic sequences includes:

[0023] Obtain the hash value by concatenating the user-defined key and the system time. The formula is as follows:

[0024] ;

[0025] Among them, is the user input key, is the system time, is the hash value after hash processing, represents the hash function, which is used to perform hash transformation on the input key and output a 256-bit hash sequence;

[0026] Input the hash value after hash processing into the following Young's hyperchaotic system for iteration:

[0027] ;

[0028] Among them, is the control variable of the system, , , are the input values for each iteration, , , are the output values for each iteration;

[0029] Ultimately, the elements of all iterations form three aperiodic chaotic sequences.

[0030] Furthermore, the three pre-generated aperiodic chaotic sequences are used to perform hierarchical encryption on three-level privacy regions, including:

[0031] For the background region, a stream key diffusion algorithm is adopted. A stream key is dynamically generated through the chaotic sequence and used to perform diffusion encryption on all image pixels;

[0032] For the face contour region, an improved Arnold transform is adopted, and high-frequency trigonometric functions are combined to perform coordinate scrambling encryption on pixel blocks;

[0033] For the face feature region, a channel-coupled chaotic scrambling algorithm is adopted to encrypt the pixel values of the RGB channels through the chaotic sequence coupling.

[0034] [[ID=;]]Further, the step of adopting a stream key diffusion algorithm for the background region, dynamically generating a stream key through a chaotic sequence, and performing diffusion encryption on all image pixels includes:

[0035] Based on the elements of the chaotic sequence generate a stream key , and the formula is as follows:

[0036] ;

[0037] where sin(10 );

[0038] Perform diffusion encryption on all image pixels through the formula ;

[0039] where represents the value of the th pixel after diffusion, respectively represent the th, th elements in the stream key sequence, ris the original pixel.

[0040] Furthermore, the step of adopting an improved Arnold transform for the face contour region and combining high-frequency trigonometric functions to perform coordinate scrambling encryption on pixel blocks includes:

[0041] The face contour area is divided into 32×32 pixel blocks, and through the dynamic parameter formula

[0042] ;

[0043] Nonlinear scrambling encryption is performed on the pixel coordinates. Among them, represents the pixel block coordinates after transformation, is the original block coordinate, is the control parameter value, is the width and height of the image.

[0044] Furthermore, the channel-coupled chaos scrambling algorithm is adopted for the face feature area, and the pixel values of the RGB channels are encrypted through the chaos sequence coupling. The formula is as follows:

[0045] ;

[0046] Among them, represents the three color channels of the picture, represents the three channels after transformation.

[0047] In a second aspect, the present invention provides a user privacy protection device for a face recognition system, including:

[0048] A memory for storing computer programs / instructions;

[0049] A processor for executing the computer programs / instructions to implement the steps of the method described in any one of the foregoing.

[0050] In a third aspect, the present invention provides a computer-readable storage medium, on which a computer program is stored, and when the program is executed by a processor, the steps of the method described in any one of the foregoing are implemented.

[0051] Compared with the prior art, the beneficial effects achieved by the present invention:

[0052] The present invention provides a user privacy protection method and device for a face recognition system, and achieves efficient and secure protection through the following innovations:

[0053] (1) Fast face recognition and privacy classification algorithm design: A fast face recognition and privacy classification algorithm based on edge detection and lightweight semantic segmentation is proposed to achieve fast recognition and positioning of the face area, and through the calculation of privacy weights, the pictures containing faces collected by the face recognition system are divided into regional blocks of different privacy levels;

[0054] (2) Strong chaotic sequence generation method: Based on the Yang's hyperchaotic system, a hash that combines user-defined keys and system time is proposed to generate collision-resistant chaotic sequences, solving the threat that traditional chaotic algorithms are vulnerable to differential attacks and providing the possibility for lightweight image encryption.

[0055] (3) Lightweight hierarchical encryption algorithm: For regions with different privacy levels, based on the designed strong privacy chaotic sequence generation method, a stream key diffusion model, a lightweight Arnold transform method, and a channel-coupled chaotic scrambling method are respectively proposed to achieve image encryption according to the privacy level and protect user privacy in face recognition.

[0056] This solution ensures the privacy of background information while solving the problems of low encryption efficiency and irreversibility in the existing technology, and is applicable to high-real-time scenarios such as face recognition and video transmission. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 is a flowchart of a user privacy protection method for a face recognition system provided by an embodiment of the present invention;

[0058] Figure 2 is a schematic diagram of the collaborative work process of each module provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0059] The present invention will be further described below with reference to the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention and should not be used to limit the protection scope of the present invention.

[0060] Embodiment 1. This embodiment introduces a user privacy protection method for a face recognition system, including:

[0061] Perform parallel processing of edge detection and lightweight semantic segmentation on the input image to generate an edge intensity map and a semantic confidence map;

[0062] Generate a privacy grading mask by dynamically weighting and fusing the edge intensity map and the semantic confidence map;

[0063] Divide the image into three levels of privacy regions through the privacy grading mask;

[0064] Perform hierarchical encryption on the three levels of privacy regions through three pre-generated aperiodic chaotic sequences to complete user privacy protection;

[0065] Among them, the three aperiodic chaotic sequences are iteratively generated by the Yang's hyperchaotic system based on the hash value concatenated with the user-defined key and the system time.

[0066] Such as Figure 1 and Figure 2As shown, the user privacy protection method for the face recognition system provided in this embodiment specifically involves the following steps in its application process:

[0067] S1. Fast face recognition and privacy grading: For the captured image by the camera, use the edge detection-semantic segmentation model to quickly and efficiently divide the entire image into three privacy levels:

[0068] (Ⅰ) Generate the face geometric contour through the edge detection modality, where the edge detection threshold is dynamically calculated based on the local statistical characteristics of the image, and the following calculation formula is designed for the dynamic fusion weight:

[0069] ；

[0070] ；

[0071] Among them, is the mean of local semantic confidence, is the standard deviation of edge intensity; is the fusion weight of semantic segmentation, is the fusion weight of edge detection;

[0072] (Ⅱ) Generate the pixel-level semantic confidence mapping through the lightweight semantic segmentation network MobileNetV3-Small, and the lightweight semantic segmentation confidence model outputs ∈[0,1];

[0073] (Ⅲ) Perform weighted fusion on the edge intensity mapping and the semantic confidence mapping to obtain a three-level region, and its fusion mask is:

[0074] ;

[0075] Among them, the semantic segmentation confidence mapping is the output value of the semantic segmentation model, ∈[0,1], and is the output of the Canny algorithm (an edge detection algorithm);

[0076] When is marked as a first-level region, 0.6 ≤ [[ID= fifty]] ≤ 0.9 is marked as a second-level region, and all pixels of the image belong to the range of the third-level region; the first-level privacy region is the region with the highest privacy level and belongs to the face feature area in the image. The feature vector values extracted from this part of the region during the face acquisition process contain all the data for biometric individual authentication and are extremely sensitive. Therefore, relatively complex compound chaotic encryption will be performed.

[0077] The secondary privacy zone is the area within the face frame. Facial images involve personal privacy, so this area has the second priority for privacy protection. Considering the balance between security and computational overhead, the improved Arnold transform (an image scrambling algorithm) can better adapt to this scenario.

[0078] The third-level privacy zone includes the aforementioned areas and the background area. The background area carries digital traces, such as scene information and landmarks, which could expose the photographer's privacy. However, considering the balance between efficiency and security of the entire system, it is marked as a lower-priority privacy zone for faster processing. All pixels are also defined as third-level privacy zones for processing, allowing them to participate in a single round of diffusion, thus avoiding the exposure of pixel statistical patterns after scrambling the first and second-level zones.

[0079] S2. Hyperchaotic sequence pre-generation. At the module design level, the chaotic sequence is run in parallel with other modules and pre-calculated to improve system efficiency. By iterating the Yang system, three hyperchaotic sequences are obtained for encryption, which are mathematically expressed as follows:

[0080]

[0081] 、 、 is the input value of each iteration, and the output 、 、 , and proceed to the next iteration.

[0082] Finally, all iterated elements form three mutually coupled chaotic sequences . Is the control parameter of the system. The initial value of the system depends on an input value from the user , and the system time at that time . and After hash concatenation, a collision-resistant hash value is obtained, which is normalized and input into the system to start iteration.

[0083] At the beginning of the implementation of the entire solution, the module will pre-calculate a sufficient number of sequences to improve efficiency. Subsequently, it will be performed asynchronously with the fast face recognition, privacy classification, and layered encryption modules through a parallel architecture.

[0084] S3, perturbation and layered encryption. After the layered encryption module receives the image to be encrypted that is marked with a privacy level:

[0085] S301 , performing random perturbations on the least significant bit plane of the image by several bits. The perturbations can be amplified by a chaotic system, thereby effectively countering differential attacks and plaintext attacks.

[0086] S302. Perform fast stream key encryption on the third-level privacy area, i.e., all pixels of the picture. The mathematical formula for generating the stream key is:

[0087] ;

[0088] Among them, is a defined non-linear combination operation, specifically sin(10 ); represents the stream key, is an element of the chaotic sequence.

[0089] The mathematical formula for stream diffusion through the stream key is:

[0090] ;

[0091] Among them, represents the value after diffusion of the th pixel, respectively represent the th, th elements in the stream key sequence, is the original pixel. Calculate , and perform a simple addition transformation between the plaintext and the key stream. Then calculate Introduce the sine function to make the key stream more complex and non-linear.

[0092] S303. For the second-level privacy area, i.e., the part within the face frame. Since this area has privacy information in the visual sense, pixel block scrambling is designed to eliminate the visual meaning of the picture. Specifically, based on the classical Arnold transformation, by introducing high-frequency trigonometric functions as coefficients, the linear characteristics of the classical transformation itself are eliminated, and at the same time, floor discretization enhances chaos. The entire transformation process adopts a block mode. According to the pixel size of the picture, a 32-pixel size block transformation is designed to achieve stronger versatility and higher efficiency. The formula for the designed lightweight transformation is as follows:

[0093] ；

[0094] where respectively represent the pixel coordinates before and after the transformation, , are control parameters, is the width and height of the picture.

[0095] S304, the primary privacy area, i.e., the facial feature area, not only contains visual meaning but can also extract individual feature values, so it has a higher privacy level. However, the computational overhead of the system needs to be compatible with low-computing-power devices, so some traditional high-strength and multi-round image encryptions are not feasible. To solve this problem, we obtain higher security performance by coupling three color channels while controlling the computational overhead. The proposed channel-coupled chaos scrambling method is as follows:

[0096] ;

[0097] respectively represent the three color channels of the original image, , , are elements in the chaos sequence generated by the chaos system, is the changed color channel.

[0098] This change couples the three color channels, enhances chaos, has good security performance, and the pixels in the primary privacy area are few, which can effectively save computing resources.

[0099] Next, in combination with a preferred embodiment, the content involved in the above embodiment will be described.

[0100] Application scenario:

[0101] Real-time face privacy protection in an intelligent access control system requires real-time encryption of face and background information in the video stream.

[0102] Hardware configuration:

[0103] 1. Main control platform: NVIDIA Jetson AGX Xavier (main control platform under NVIDIA): GPU: 512-core Volta architecture (NVIDIA GPU micro-architecture), CPU: 8-core ARM v8.2 64-bit, memory: 16GB LPDDR4x.

[0104] 2. Camera module: Sony IMX477 (model) sensor, supports 1920×1080@30fps input, ROI (region of interest) programmable configuration.

[0105] 3. Encryption acceleration: Use the CUDA cores (NVIDIA GPU parallel computing units) of the main control platform for parallel computing of chaos sequences and image scrambling.

[0106] 4. Software environment:

[0107] a) Operating system: Ubuntu 20.04 LTS (Ubuntu operating system).

[0108] b) Deep learning framework: PyTorch 1.10 deep learning framework + TensorRT 8.2 inference acceleration engine.

[0109] c) Encryption library: OpenSSL 3.0 open-source encryption toolkit (used for SHA3-256 hash calculation).

[0110] Implementation steps:

[0111] S1. Fast face recognition and privacy grading;

[0112] S101. Implementation of edge detection modality: Use improved Canny edge detection and optimize the threshold adaptive mechanism. Its parameter configuration is as follows:

[0113] Gaussian filtering: Kernel size 5×5, standard deviation , suppressing high-frequency noise.

[0114] Gradient calculation: Sobel operator (edge detection operator, horizontal kernel [−1,0,1;−2,0,2;−1,0,1], vertical kernel transposed)

[0115] Dynamic double threshold: High threshold , low threshold where and are the grayscale mean and standard deviation of the local 8×8 region.

[0116] Edge connection: Adopt the bidirectional scanning method to preferentially connect strong edge pixels.

[0117] Output: Edge intensity map Sedge∈[0,1], Sedge∈[0,1], normalized to a floating-point matrix.

[0118] S102. Implementation of semantic segmentation modality, using lightweight MobileNetV3-Seg (lightweight image segmentation network), with the following optimizations:

[0119] Backbone network: MobileNetV3-Small (miniaturized network model), the number of depthwise separable convolutional layers reduced to 8 layers, with 1.2M parameters.

[0120] Decoder design: Adopt transposed convolutional operation with a 3×3 convolutional kernel and stride = 2, and gradually upsample to the original resolution.

[0121] Simplification of the Atrous Spatial Pyramid Pooling (ASPP) module: Only retain 3 parallel branches (atrous rate = 6, 12, 18), and compress the output channels to 64.

[0122] Output layer: Sigmoid activation function generates pixel-level semantic confidence map ∈[0,1].

[0123] Dataset: CelebAMask-HQ dataset (30,000 face images with facial features, contours, and background annotations).

[0124] Loss function: Use Combined loss function Used to solve class imbalance.

[0125] Training parameters: Adam optimizer (Adaptive Moment Estimation), lr=1e-4, batch_size=16.

[0126] S103, dynamic weight fusion and region labeling;

[0127] Fusion formula:

[0128] ;

[0129] ;

[0130] Implementation steps:

[0131] Local statistics calculation: Divide the image into 32×32 blocks and calculate each block (mean semantic confidence) and (Standard deviation of edge strength).

[0132] High semantic confidence zone ( >0.7): Increase the α weight to 0.7-0.9.

[0133] High edge complexity area ( >0.4): Increase the β weight to 0.6-0.8.

[0134] Area marking rules:

[0135] First-level area: >0.9, covers facial features (eyes, nose, mouth), sample size 80×80 pixels.

[0136] Secondary area: 0.6≤ ≤0.9, covering the face contour (example 320×320 pixels).

[0137] Level 3 area: All pixels in the entire image are forced to participate in the lowest intensity encryption.

[0138] Post-processing: Perform 3×3 median filtering on adjacent blocks to eliminate regional fragmentation.

[0139] S2, hyperchaotic sequence generation and key management;

[0140] S201, user key processing;

[0141] User key K: string format, length 8-64 characters, allowed character set is [A-Za-z0-9@#$%^&+=].

[0142] System time T: millisecond-level precision UTC timestamp (Coordinated Universal Time timestamp), in the Unix Epoch format.

[0143] Hash concatenation: Call the EVP_sha3_256 function of the OpenSSL library (the calculation interface provided by the encryption library) to output a 64-byte hash value.

[0144] Initial value generation:

[0145] The hash value It is divided into three parts:

[0146] ；

[0147] ;

[0148] ;

[0149] Normalized to the interval [−1,1].

[0150] S202, Yang's hyperchaotic system iteration;

[0151] Kinetic equation:

[0152] ;

[0153] Through Lyapunov (Lyapunov index) analysis ( =0.52, =0.18, = -1.02), ensuring that the system is in a hyperchaotic state.

[0154] Control parameters

[0155] Determined by bifurcation graph optimization. Pre-generate 10,000 chaotic values and store them in a ring buffer.

[0156] S3, layered encryption process;

[0157] S301 Level 3 area encryption;

[0158] Stream key generation:

[0159] ;

[0160] Encryption operation:

[0161] ;

[0162] Pixel block division: The image is divided into 64×64 blocks, and each block is assigned an independent chaotic sequence segment to avoid repeated keys.

[0163] Parallel computing: CUDA is used to accelerate the stream key generation and XOR operation, with a throughput of 1.2 GPixel / s.

[0164] S4. Secondary region encryption (improved Arnold transform);

[0165] Transformation formula:

[0166] ；

[0167] Parameter dynamic adjustment:

[0168] ;

[0169] is the value of the chaotic sequence.

[0170] Iteration times: Iteration times = ;

[0171] Block scrambling: The secondary region (320×320) is divided into 16×16 sub-blocks, and the transformation is executed in parallel.

[0172] S5. Primary region encryption (channel-coupled chaotic scrambling);

[0173] Channel coupling formula:

[0174] ;

[0175] Quantization noise injection: Items such as introduce chaotic quantization noise to destroy pixel correlation.

[0176] GPU acceleration: Texture memory caches RGB channel data, and the single-pixel calculation time ≤ 5 ns.

[0177] The output image at this time is a secure image after three rounds of hierarchical encryption.

[0178] Embodiment 2. This embodiment provides a user privacy protection device for a face recognition system, including:

[0179] A memory for storing computer programs / instructions;

[0180] A processor for executing the computer programs / instructions to implement the steps of the method according to any one of Embodiment 1.

[0181] Embodiment 3 provides a computer-readable storage medium having a computer program stored thereon, and when the program is executed by a processor, it implements the steps of the method according to any one of Embodiment 1.

[0182] The above are only the preferred embodiments of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the technical principles of the present invention, several improvements and modifications can be made, and these improvements and modifications should also be regarded as the protection scope of the present invention.

[0183] Those skilled in the art should understand that the embodiments of the present disclosure can be provided as a method, a system or a computer program product. Therefore, the present disclosure can take the form of a complete hardware embodiment, a complete software embodiment or an embodiment combining software and hardware aspects. Moreover, the present disclosure can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk memories, CD-ROMs, optical memories, etc.) containing computer-usable program codes.

[0184] The present disclosure is described with reference to the flowcharts and / or block diagrams of methods, apparatuses (systems) and computer program products according to the embodiments of the present disclosure. It should be understood that each flow and / or block in the flowchart and / or block diagram can be implemented by computer program instructions, and the combination of the flows and / or blocks in the flowchart and / or block diagram can also be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing devices to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing devices generate a device for implementing the functions specified in Figure 1 one or more flows and / or blocks Figure 1 one or more blocks.

[0185] These computer program instructions can also be stored in a computer-readable memory that can guide a computer or other programmable data processing devices to work in a specific manner, so that the instructions stored in the computer-readable memory generate a manufactured product including an instruction device, and the instruction device implements the functions specified in Figure 1 one or more flows and / or blocks Figure 1 one or more blocks.

[0186] These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus, so that a series of operation steps are performed on the computer or other programmable apparatus to produce a computer-implemented process, thereby providing instructions for implementing the processes Figure 1 in one process or multiple processes and / or boxes Figure 1 steps of the functions specified in one box or multiple boxes.

[0187] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present disclosure rather than limit the scope of its protection. Although the present disclosure has been described in detail with reference to the above embodiments, those of ordinary skill in the art should understand that: after reading the present disclosure, those skilled in the art may still make various changes, modifications or equivalent replacements to the specific implementation manners of the invention, but these changes, modifications or equivalent replacements are all within the scope of protection of the pending claims of the disclosure.

Claims

1. A user privacy protection method for a face recognition system, characterized in that: include: Perform edge detection and lightweight semantic segmentation on the input image in parallel to generate edge strength map and semantic confidence map; Generate privacy-graded masks by fusing edge strength maps and semantic confidence maps through dynamic weighting; including: Based on the mean of local semantic confidence and the standard deviation of edge strength , calculate the fusion weight of semantic segmentation and edge detection, the formula is as follows: ； in, is the fusion weight of semantic segmentation, is the fusion weight of edge detection; Through the weighted formula Generate a fusion mask where is the semantic confidence mapping, is the edge strength map; The image is divided into three levels of privacy areas through privacy classification masks; including: According to the fusion mask The threshold range falls into which the image is divided into three levels of privacy areas: facial feature area, facial contour area and background area; Through the three pre-generated periodic chaotic sequences, the three-level privacy areas are hierarchically encrypted to achieve user privacy protection; The three period-free chaotic sequences are generated iteratively through the Yang's hyperchaotic system based on the hash value of the user-defined key and the system time; including: The hash value is obtained by concatenating the user-defined key and the system time. The formula is as follows: ； in, is the user-entered key, is the system time, is the hash value after hashing. Indicates that the generated hash value length is 256 bits; The hash value after hashing , input the following Yang's hyperchaotic system for iteration: ； in, is the control variable of the system, 、 、 is the input value for each iteration, 、 、 is the output value of each iteration; Finally, all iterated elements form three periodic chaotic sequences.

2. The user privacy protection method for a face recognition system according to claim 1, characterized in that: The three levels of privacy areas are hierarchically encrypted using three pre-generated periodic chaotic sequences, including: A stream key diffusion algorithm is used for the background area, which dynamically generates stream keys through chaotic sequences and diffuses and encrypts all pixels in the image. The improved Arnold transform is used for the face contour area, and the high-frequency trigonometric function is combined to scramble the coordinates of the pixel blocks for encryption. A channel-coupled chaotic scrambling algorithm is used for facial feature areas, and the RGB channel pixel values are encrypted by coupling the chaotic sequence.

3. The user privacy protection method for a face recognition system according to claim 2, characterized in that: The stream key diffusion algorithm is used for the background area, the stream key is dynamically generated through the chaotic sequence and the diffusion encryption is performed on the pixels of the entire image, including: Based on chaotic sequence elements Generate stream key , the formula is as follows: ； in sin(10 ) ; By formula Diffusion encryption is performed on all pixels of the image; in, Indicates the The value of the pixel after diffusion, Respectively represent the first 、 elements, It is the original pixel.

4. The user privacy protection method for a face recognition system according to claim 3, characterized in that: The improved Arnold transform is used for the face contour area, and high-frequency trigonometric functions are combined to perform coordinate scrambling encryption on pixel blocks, including: The face contour area is divided into 32×32 pixel blocks, and the dynamic parameter formula is used to calculate the face contour area. ； Perform nonlinear scrambling encryption on pixel coordinates, where represents the transformed pixel block coordinates, are the original block coordinates, is the control parameter value, is the width and height of the image.

5. The user privacy protection method for a face recognition system according to claim 4, characterized in that: The channel-coupled chaotic scrambling algorithm is used for the facial feature area, and the chaotic sequence is coupled with the RGB channel pixel value encryption. The formula is as follows: ； Among them, Represents the three color channels of the image, It represents the three channels after transformation.

6. A user privacy protection device for a face recognition system, characterized in that: include: Memory, used to store computer programs / instructions; A processor configured to execute the computer program / instructions to implement the steps of the method according to any one of claims 1 to 5.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the method according to any one of claims 1 to 5 are implemented.

Images