A management method based on campus information security

Abstract

This invention relates to a management method based on campus information security, comprising: acquiring the student's original information image; splitting the original information image into several original pixel blocks based on an image segmentation mechanism; constructing a multidimensional noise model; extracting corresponding noise data from the multidimensional noise model and injecting it into the corresponding original pixel blocks; re-merging the injected original pixel blocks and smoothing the boundaries between adjacent pixel blocks to form a new face image; storing the new face image; using a noise injection method to modify each original pixel block, eliminating the original real image features, and forming a fake face image, thereby ensuring the security of real data and reducing the attacker's motivation to crack the image; and making the attacker assume that the fake face image obtained by the attack is a real face image, thereby further improving the security of the student's original information image.

Description

Technical Field

[0001] This invention relates to the field of information data security technology, and in particular to a management method based on campus information security. Background Technology

[0002] The construction of smart campuses has become a key task of concern for many universities. The risks and challenges to cybersecurity are also becoming increasingly serious. A large amount of student image data is stored in the smart campus system, and these stored images are usually not processed or encrypted. Therefore, once a network attack occurs, it is very likely that these unprocessed image data in the system will be leaked or tampered with, which will lead to the infringement of students' rights and personal privacy.

[0003] Chinese invention patent application number 202210850581.2 provides a smart campus network security protection system, which converts the grayscale values ​​of each pixel in the original information image into a binary sequence; obtains the bit block corresponding to each pixel based on the binary sequence, the bit block is composed of sub-bit blocks with values ​​and sub-bit blocks without values; obtains the target identification code sequence based on the bit block and the kernel value in the target filtering kernel; obtains the target bit block based on the target identification code sequence; obtains the value of each sub-pixel block in the pixel block corresponding to each pixel based on the target bit block and the target filtering kernel; scrambles each sub-pixel block in each pixel block to obtain the target pixel block corresponding to each pixel, and stores the target pixel block corresponding to each pixel.

[0004] However, the scrambled image data is significantly different from the original image data, with obvious signs of modification, which to some extent increases the motivation of network attackers. Furthermore, the random scrambling of the system still has fixed disturbance patterns and paths, which limits the direction of cracking and thus reduces the security of the image data. Summary of the Invention

[0005] This invention addresses the technical problems existing in the prior art by providing a management method based on campus information security.

[0006] The technical solution of the present invention to solve the above-mentioned technical problems is as follows:

[0007] A management method based on campus information security, the method comprising:

[0008] S100: Obtain the student's original information image and divide the original information image into several original pixel blocks based on the image block mechanism;

[0009] S200: Construct a multi-dimensional noise model, classify different levels based on the content and sensitivity of different original pixel blocks, and set corresponding noise type parameters and intensity parameters for different levels of original pixel blocks; extract the corresponding noise data from the multi-dimensional noise model and inject it into the corresponding original pixel blocks according to the corresponding noise type parameters and intensity parameters;

[0010] S300: Re-fuse the injected original pixel blocks and smooth the boundaries between adjacent pixel blocks to form a new face image, and store the new face image;

[0011] S400: Perform quality assessment on the new face image to obtain an assessment value. If the assessment value is greater than the preset assessment threshold, store the corresponding new face image. If the assessment value is not greater than the preset assessment threshold, repeat steps S300 and S400 until the assessment value is greater than the preset assessment threshold or the preset maximum number of optimizations is reached.

[0012] The method for constructing a multidimensional noise model includes: determining the model's dimensions and initial parameters; selecting a chaotic system to generate noise data; distributing the generated noise data to the various dimensions of the model according to injection rules; and combining the distributed noise data to form a multidimensional noise model. Further, the method also includes:

[0013] S500: Identify occluded parts of the face and parts with changes in expression, and mark them as noise-free raw pixel blocks;

[0014] S600: Establish a random face image library, calculate the image matching degree between the original information image and all face images in the random face image library, and select the random face image with the highest image matching degree as the interference image; based on the image block mechanism, split the interference image into several interference pixel blocks; the interference pixel blocks correspond one-to-one with the original pixel blocks; continue to execute step S200;

[0015] S700: Select the interfering pixel block corresponding to the noiseless original pixel block, re-fuse the interfering pixel block and the injected original pixel block, and smooth the boundary between adjacent pixel blocks to form a new face image; continue to execute step S400.

[0016] Furthermore, step S100 also includes:

[0017] S110: Detect the surface features of the original information image and obtain the feature value. If the feature value is greater than the preset feature threshold, proceed to step S500; if the feature value is not greater than the preset feature threshold, proceed to step S200.

[0018] Furthermore, the surface feature recognition includes facial region recognition, facial key point detection, facial occlusion and expression change classification, and provides corresponding factor values ​​based on the recognition results; the feature value is a comprehensive value obtained by weighted summation of all factor values.

[0019] Furthermore, in step S300, the boundaries between adjacent pixel blocks are smoothed, including: smoothing using Gaussian filtering, as detailed below:

[0020]

[0021] in, This represents the smoothed pixel value; This represents the pixel value at coordinates (x, y) in the image; i and j are the horizontal and vertical offsets in the Gaussian filter kernel, respectively; the values ​​of i and j range from [-k, k], where k is the radius of the filter. This refers to the Gaussian kernel function, and the specific formula is as follows:

[0022]

[0023] Where σ is the standard deviation of the Gaussian kernel, which controls the smoothness of the filter; It represents the square of the distance from the neighboring pixel (i, j) to the center pixel (0, 0).

[0024] Further, in step S400, the evaluation metrics for quality assessment include image sharpness, noise level, and information retention; specifically, peak signal-to-noise ratio (PSNR) is used for evaluation, as follows:

[0025]

[0026] in, It is an evaluation value. It is the square of the maximum pixel value of the image. This is the mean squared error, calculated as follows:

[0027]

[0028] Where M and N are the height and width of the image, respectively.

[0029] The beneficial effects of this invention are: by using noise injection to modify each original pixel block, the original real image features are eliminated, and a fake face image is formed, which achieves the effect of ensuring the security of real data and reducing the attacker's motivation to crack; it makes the attacker assume that the fake face image obtained by the attack is a real face image, thereby further improving the security of students' original information images. Attached Figure Description

[0030] Figure 1 This is a schematic diagram of a management method based on campus information security in an embodiment of the present invention. Figure 1 ;

[0031] Figure 2 This is a schematic diagram of a management method based on campus information security in an embodiment of the present invention. Figure 2 . Detailed Implementation

[0032] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0033] In the description of this application, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the stated features. In the description of this application, "multiple" means two or more, unless otherwise explicitly specified.

[0034] In the description of this application, the term "for example" is used to mean "used as an example, illustration, or description." Any embodiment described as "for example" in this application is not necessarily to be construed as being more preferred or advantageous than other embodiments. The following description is provided to enable any person skilled in the art to make and use the invention. Details are set forth in the following description for purposes of explanation. It should be understood that those skilled in the art will recognize that the invention can be made without using these specific details. In other instances, well-known structures and processes will not be described in detail to avoid obscuring the description of the invention with unnecessary detail. Therefore, the invention is not intended to be limited to the embodiments shown, but is consistent with the broadest scope of the principles and features disclosed in this application.

[0035] Example 1: As Figure 1 As shown, a management method based on campus information security includes:

[0036] S100: Obtain the student's original information image and divide the original information image into several original pixel blocks based on the image block mechanism;

[0037] In some embodiments, the student's original information images are obtained through a student information management system. These images typically contain sensitive data such as the student's facial features and identity information. The obtained original images are preprocessed, including operations such as denoising, grayscale conversion, and normalization, to ensure the accuracy of subsequent processing. This is a well-known prior art and will not be described in detail here.

[0038] The image segmentation mechanism is as follows: the original image is converted to grayscale, the facial region is located using a facial detection algorithm, and the facial region is initially divided into several large pixel blocks based on facial feature points; the importance of each large pixel block is evaluated, and the large pixel blocks are classified into first-level and second-level according to the importance evaluation results; the first-level large pixel blocks are further divided equally based on the distribution, shape, and texture of feature points; the pixel blocks formed by the initial division and the second equal division are integrated to obtain several original pixel blocks.

[0039] The importance assessment is a comprehensive evaluation based on the location of large pixel blocks on the face, the feature information they contain, and personalized features. Preferably, among the facial features of each person, the eyes, nose, and mouth are considered more important. Different people also exhibit personalized features on their faces, such as the presence of obvious moles or birthmarks. When conducting the importance assessment, a comprehensive evaluation should be conducted based on each person's personalized features, and image data of all people should not be processed in a homogenized manner.

[0040] Taking a human face image as an example, the specific segmentation steps are as follows:

[0041] Preliminary segmentation: The facial region is located using a facial detection algorithm, and the facial region is divided into four large pixel blocks: eye region, nose region, mouth region, and ear region based on facial feature points.

[0042] The importance of each large pixel block is evaluated separately: the eye and nose regions are given the highest importance because they play a key role in facial recognition, while the mouth and ear regions are given lower importance, but their feature information still needs to be considered.

[0043] The eye region is further divided into several smaller pixel blocks (assuming the large pixel block size is...). The size of each smaller pixel block is The image is then divided into (Each pixel block contains the left and right eyes respectively.)

[0044] The nose region is further subdivided into several pixel blocks based on the features of the bridge and alar. The results of the initial and secondary subdivisions are then integrated to form a final subdivision containing multiple pixel blocks. Each pixel block has a clear boundary and identifier.

[0045] S200: Construct a multi-dimensional noise model, classify different levels based on the content and sensitivity of different original pixel blocks, and set corresponding noise type parameters and intensity parameters for different levels of original pixel blocks; extract the corresponding noise data from the multi-dimensional noise model and inject it into the corresponding original pixel block according to the corresponding noise type parameters and intensity parameters.

[0046] In some embodiments, pixel blocks are divided into different levels based on their content (e.g., facial features, background, etc.) and sensitivity (e.g., whether they contain identity information, whether they belong to unique personal characteristics, etc.). For example, high-sensitivity pixel blocks, medium-sensitivity pixel blocks, and low-sensitivity pixel blocks. Noise types include Gaussian noise, salt-and-pepper noise, and speckle noise, etc. Different types of noise have different effects on the image. Gaussian noise is suitable for areas requiring smoothing, and its intensity controls the amplitude of the noise. Salt-and-pepper noise is suitable for areas requiring random point noise, and its intensity controls the density of the noise points. Speckle noise is suitable for areas requiring simulated natural noise, and its intensity controls the density of the noise. The specific noise type should be selected according to the actual situation, and this application does not impose specific limitations. Noise intensity determines the degree to which noise occludes the image; the greater the intensity, the more obvious the occlusion effect, but it also has a greater impact on the visual quality of the image. Based on the set noise type and intensity parameters, the corresponding noise data is extracted from the multidimensional noise model and injected into the corresponding original pixel block.

[0047] S300: The injected original pixel blocks are re-fused, and the boundaries between adjacent pixel blocks are smoothed to form a new face image;

[0048] In some embodiments, the pixel blocks injected with noise are reassembled into a complete image. Since each pixel block is processed independently, this may result in discontinuous boundaries between blocks. Smoothing the boundaries between adjacent pixel blocks reduces block artifacts. Gaussian filtering, mean filtering, etc., can be used here. The specific calculation formula for Gaussian filtering is as follows:

[0049]

[0050] in, This represents the smoothed pixel value; This represents the pixel value at coordinates (x, y) in the image; i and j are the horizontal and vertical offsets in the Gaussian filter kernel (also known as the convolution kernel), respectively; the values ​​of i and j range from [-k, k], where k is the radius of the filter. This refers to the Gaussian kernel function, and the specific formula is as follows:

[0051]

[0052] Here, σ is the standard deviation of the Gaussian kernel, which controls the smoothness of the filter. The larger σ is, the stronger the smoothing effect. This represents the square of the distance from a neighboring pixel (i, j) to the center pixel (0, 0). The value of the Gaussian kernel function decreases as the distance increases, so the neighboring pixels farther from the center pixel contribute less to the smoothing result.

[0053] S400: Perform quality assessment on the new face image to obtain an evaluation value. If the evaluation value is greater than the preset evaluation threshold, store the corresponding new face image. If the evaluation value is not greater than the preset evaluation threshold, repeat steps S300 and S400 until the evaluation value is greater than the preset evaluation threshold or the preset maximum number of optimizations is reached.

[0054] In some embodiments, constructing a multidimensional noise model specifically includes: determining the dimensions and initial parameters of the noise model, such as the mean, variance, and frequency distribution of the noise, based on the needs of image processing, which will affect the characteristics and distribution of the generated noise data; selecting a chaotic system to generate noise data based on the complexity and randomness requirements of the noise data, since chaotic systems have good randomness and unpredictability and can generate noise data that meets the requirements; using the selected chaotic system, generating a noise data sequence according to the set initial conditions and parameters; preprocessing the generated noise data sequence, such as normalization and filtering, to ensure that the characteristics and distribution of the noise data meet the requirements of the model; formulating noise data injection rules based on the goals and needs of image processing, the injection rules should consider factors such as the distribution, intensity, and frequency of the noise data, as well as the protection and enhancement of image features; allocating the generated noise data to each dimension of the model according to the injection rules; adjusting the intensity of the noise data injected into each dimension according to the needs of image processing, the noise intensity is achieved by adjusting the amplitude, frequency, or distribution of the noise data; and combining the allocated noise data dimensions to form a multidimensional noise model.

[0055] In some embodiments, the evaluation metrics for quality assessment include image sharpness, noise level, and information retention. This application uses Peak Signal-to-Noise Ratio (PSNR) for evaluation. Before evaluation, a benchmark value needs to be set based on historical experimental data. Since this application adds noise to the image, a practical evaluation threshold needs to be set when using this evaluation method. PSNR is an image quality evaluation metric calculated based on MSE; a higher value indicates better image quality. Its unit is decibels (dB), and it is used to measure the performance of image processing algorithms, as detailed below:

[0056]

[0057] in, It is an evaluation value. It is the square of the maximum pixel value of the image. This is the mean squared error, which is the average of the squared differences in pixels between the original image and the processed image. The specific calculation is as follows:

[0058]

[0059] Where M and N are the height and width of the image, respectively.

[0060] In some embodiments, the evaluation value is compared with a preset evaluation threshold. If the evaluation value is greater than the threshold, the image quality is considered acceptable, and the new face image is stored; otherwise, the process returns to step S300 for further optimization. A maximum number of optimization attempts is also set. If the evaluation threshold is not met even after reaching the maximum number of optimization attempts, optimization is stopped, and the current image is output.

[0061] The technical solutions described in the embodiments of this application have at least the following technical effects or advantages:

[0062] This application uses noise injection to modify each original pixel block, eliminating the original real image features and forming a fake face image. This achieves the effect of ensuring the security of real data and reducing the attacker's motivation to crack the image. It also makes the attacker assume that the fake face image obtained by the attack is a real face image, thereby further improving the security of the student's original information image.

[0063] Example 2: In the above example, noise is injected into the student's original information image to generate a new fake face image for storage. However, when performing noise injection, if the original information image has facial occlusion or expression changes, the effect of noise injection is limited because some facial features have been occluded or changed. The generated image may still retain some features of the original image, which reduces the difficulty of cracking.

[0064] Step S100 also includes:

[0065] S110: Detect surface features of the original information image and obtain feature values. If the feature value is greater than the preset feature threshold, proceed to step S500; if the feature value is not greater than the preset feature threshold, proceed to step S200.

[0066] The surface feature recognition includes facial region recognition, facial key point detection, facial occlusion and expression change classification, and provides corresponding factor values ​​based on the recognition results; the feature value is a comprehensive value obtained by weighted summation of all factor values;

[0067] In some embodiments, a face detection algorithm (such as a deep learning algorithm) is used to identify the facial region in the original information image and output the coordinates and size of the facial region; within the identified facial region, a key point detection algorithm is used to locate the key points of the face, such as the eyes, nose, and mouth, and output the coordinates of the key points. The same detection method as in Embodiment 1 above is used here.

[0068] In some embodiments, a classification algorithm is used to determine whether a face is occluded and the type of facial expression change (such as smiling, angry, etc.), and the classification results of occlusion and facial expression change are output. Specifically, a dataset of face images containing different occlusion situations (such as masks, hats, hand occlusion, etc.) and facial expression changes (such as smiling, angry, surprised, etc.) is collected. The dataset is labeled to indicate whether each image is occluded and the type of expression, and training, validation, and test sets are divided. A convolutional neural network architecture is selected, and a network model is built, including convolutional layers, pooling layers, fully connected layers, etc., and appropriate activation functions and loss functions (such as cross-entropy loss) are set.

[0069] Train the CNN model using the prepared training set. During training, use optimization algorithms (such as Adam, SGD, etc.) to tune model parameters and minimize the loss function. Evaluate model performance using the validation set and adjust hyperparameters (such as learning rate, batch size, etc.) to improve model accuracy. Test the trained CNN model using the test set to evaluate its performance on occlusion and facial expression change classification tasks. Calculate metrics such as accuracy, recall, and F1 score to ensure the model meets practical requirements.

[0070] Using a trained CNN model, input the face image to be detected, and output the classification results of occlusion and expression changes, such as "unoccluded + smiling" and "occluded (mask) + angry".

[0071] In some embodiments, corresponding factor values ​​are assigned based on the classification results of facial region recognition, keypoint detection, occlusion, and expression changes. Facial region recognition factor value: A score is assigned based on the accuracy and completeness of facial region recognition. For example, complete recognition with clear boundaries is assigned 10 points, while partial recognition or blurred boundaries are assigned 6 points. Keypoint detection factor value: A score is assigned based on the accuracy and number of keypoints detected. For example, accurate detection of all keypoints is assigned 10 points, while missing or misaligned keypoints are assigned lower scores. Occlusion factor value: A score is assigned based on the severity and type of occlusion. For example, no occlusion is assigned 10 points, slight occlusion (such as partial eye occlusion) is assigned 8 points, and severe occlusion (such as full face occlusion) is assigned 2 points. Expression change factor value: A score is assigned based on the type and significance of expression changes. For example, natural expressions (such as smiling) are assigned 10 points, significant expression changes (such as crying) are assigned 8 points, and unnatural expressions (such as contortions) are assigned lower scores.

[0072] The feature value is calculated using a weighted summation formula: Feature value = w1 × Facial region recognition factor value + w2 × Key point detection factor value + w3 × Occlusion factor value + w4 × Expression change factor value, where w1, w2, w3, and w4 are weight coefficients, which are set according to actual needs and are not specifically limited in this application.

[0073] In some embodiments, the feature threshold can be set using various methods. It can be set based on historical experimental data and actual needs, or it can be set based on data distribution. For example, a subset of the dataset can be used to calculate the distribution of feature values, and a reasonable threshold T can be set based on statistical measures such as the mean and standard deviation of the feature values. For instance, T can be set to the mean plus a certain multiple of the standard deviation to cover the feature value range of most normal images. Specific adjustments and settings need to be made based on the actual situation and experimental results.

[0074] In some embodiments, such as Figure 2 As shown, the method further includes: S500: identifying facial occlusion parts and expression change parts, and marking them as noise-free original pixel blocks;

[0075] S600: Establish a random face image library, calculate the image matching degree between the original information image and all face images in the random face image library, and select the random face image with the highest image matching degree as the interference image; based on the image block mechanism, split the interference image into several interference pixel blocks; the interference pixel blocks correspond one-to-one with the original pixel blocks; continue to execute step S200;

[0076] S700: Select the interfering pixel block corresponding to the noiseless original pixel block, re-fuse the interfering pixel block and the injected original pixel block, and smooth the boundary between adjacent pixel blocks to form a new face image; continue to execute step S400.

[0077] In some embodiments, a large number of random face images are collected to form a random face image library; for each face image in the original information image and the random face image library, an image matching algorithm (such as structural similarity SSIM or eigenvector distance) is used to calculate the image matching degree, and the eigenvector distance is used to calculate the image matching degree, as follows:

[0078]

[0079] in, and These are the i-th feature vectors in the original image and the random image, respectively, and n is the total number of feature vectors;

[0080] Based on image matching degree, the random face image with the highest image matching degree is selected as the interference image.

[0081] In some embodiments, the original information image and the interference image are split in the same way, that is, the size of the split pixel blocks and the corresponding part allocation are the same. Facial occlusion and expression change parts are identified and marked as noise-free original pixel blocks. Based on the interference image, the parts corresponding to the noise-free original pixel blocks are found, i.e., the corresponding interference pixel blocks. The interference pixel blocks and the injected original pixel blocks are then re-fused, using a completely fake face image to replace the facial occlusion and expression change parts in the original image, and the boundaries between adjacent pixel blocks are smoothed to form a new face image; step S400 continues. In this embodiment, the formed new face image completely discards the features of the real face image.

[0082] In this embodiment, by setting up a random face image library, selecting corresponding pixel blocks to replace the original real pixel blocks, and finally fusing them, the security of face images is further improved.

[0083] Example 3: In this example, further improvements are made based on the above examples.

[0084] Step S700 also includes:

[0085] S710: After re-fusing the interfering pixel blocks and the injected original pixel blocks, an initial image is formed; the deviation type is determined based on the surface feature analysis of the interfering pixel blocks and the injected original pixel blocks, and the deviation value between each pixel block in the initial image is calculated according to the deviation type;

[0086] S720: Select the corresponding face image from the random face image library according to the deviation value and mark it as the overlay image. Superimpose the overlay image on the initial image to form a new face image.

[0087] In some embodiments, the surface feature analysis includes color distribution histogram, texture pattern recognition, and edge detection. It acquires features such as color statistics, texture direction and intensity, and edge position and shape of the image.

[0088] Based on the color distribution histogram, if the image is too dark or too bright overall, choose color deviation; if the texture direction is obvious and needs to be changed, choose texture deviation; if the position of objects in the image needs to be fine-tuned, choose position deviation.

[0089] For color deviation, set the color offset amount, such as increasing / decreasing the red channel by X units, increasing / decreasing the green channel by Y units, and keeping the blue channel unchanged or adjusting it; for texture deviation, set texture transformation parameters, such as rotation angle, scaling ratio, etc.; for position deviation, set the pixel displacement amount, such as moving M pixels horizontally and N pixels vertically.

[0090] In some embodiments, if more than one type of deviation is generated, it is necessary to perform comprehensive statistics to form a deviation value, and calculate the deviation value between each pixel block in the initial image. The specific calculation formula is as follows:

[0091]

[0092] in, It is the deviation value. This represents the weighting coefficient for the k-th type of deviation. This represents the deviation value for the k-th deviation type. This indicates the total number of deviation types.

[0093] In some embodiments, selecting a corresponding face image from a random face image library based on the deviation value and marking it as the overlay image specifically includes: combining the calculated deviation values ​​to form a deviation value feature vector; extracting corresponding feature vectors from the face images in the random face image library to form a random face feature vector; calculating the similarity between the deviation value feature vector and the random face feature vector using a similarity measurement method, and selecting the random face image with the highest similarity as the overlay image.

[0094] In some embodiments, an overlay image is superimposed on an initial image to form a new face image. Based on actual needs and circumstances, algorithms such as weighted averaging, Laplacian pyramid fusion, and Poisson fusion are dynamically selected for image fusion. This application does not impose specific limitations here. This application uses the Poisson fusion algorithm for image fusion, specifically including: selecting the region of the face image to be fused and the target region of the initial image; solving the Poisson equation to ensure a smooth transition between the fusion region and the target region at the boundary; and constructing a sparse Laplacian matrix during discretization, which describes the adjacency relationship and gradient information between image pixels. Before solving the Poisson equation, the boundary values ​​of the fusion region need to be specified. The Poisson equation is solved using linear algebraic methods (such as the conjugate gradient method), and the result is the new value of each pixel within the fusion region. The pixel values ​​obtained from the fusion region are applied to the initial image, replacing the original pixel values ​​in the target region, ensuring a natural transition between the face image and the target region without obvious boundaries.

[0095] In this embodiment, based on image deviation technology, by accurately calculating and introducing purposeful deviation values, the data layer is changed to achieve arbitrary superposition of face images, thereby enhancing the randomness and richness of the interference images and preventing attackers from discovering the protection pattern.

[0096] In this embodiment, image deviation technology is used to generate diverse facial images to avoid a single interference pattern; it ensures that the generated new facial images remain visually natural, avoiding obvious distortion or abnormality and maintaining visual naturalness; through changes at the data level, the generated facial images are made difficult for biometric systems to recognize, thus compromising biometric capabilities; for facial occlusion areas, reasonable deviation values ​​are generated to ensure the consistency of the overall protection effect.

[0097] Example 4: In the above examples, the same feature areas will still be generated in the large number of new face images generated, reducing the difficulty of cracking. This example is a further improvement on the above examples.

[0098] The method further includes:

[0099] S800: Acquire several new face images, and identify and extract feature regions with a repetition rate greater than the repetition threshold, and mark them as easily repeated regions;

[0100] S810: Abstracts the areas prone to repetition, fills them with different character codes, stores the corresponding character codes, and forms a new face image.

[0101] In some embodiments, the easily duplicated regions are abstracted, specifically including: grayscale processing of the easily duplicated regions, feature extraction of the processed easily duplicated regions, and mapping the extracted region features onto a pre-set character set to generate corresponding character codes. Mapping rules need to be pre-set to ensure that different features can be mapped to different characters, and the actual image of the easily duplicated regions is obtained based on the mapping rules.

[0102] In this embodiment, a large number of historical new face images are acquired, and regions prone to repetition are identified in the images. Specifically, feature regions in the images are extracted, and the repetition rate of these feature regions is calculated. Regions with a repetition rate greater than a preset repetition threshold are marked as prone to repetition. The repetition threshold needs to be dynamically set based on actual conditions and historical experimental data; this application does not impose specific limitations on it.

[0103] In this embodiment, the easily duplicated regions are converted to grayscale to obtain grayscale-enhanced easily duplicated regions. Features of the grayscale-enhanced easily duplicated regions are extracted to obtain region feature data. According to a pre-set character set and mapping rules, the region feature data is mapped onto the character set to generate corresponding character codes. The generated character codes are used to fill the easily duplicated regions to obtain filled easily duplicated regions. The coordinate information of the filled easily duplicated regions is obtained to determine their position in the new face image. The filled easily duplicated regions are superimposed on the new face image to replace the easily duplicated regions in the new face image. The replaced image is smoothed to eliminate edge discontinuities. The abstracted new face image is saved, completing the image processing flow. For easily duplicated regions, abstract character codes are saved instead of images, further increasing the security of the real image data.

[0104] It should be noted that the descriptions of each embodiment in the above embodiments have different focuses. For parts that are not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0105] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0106] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, as well as combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded computer, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0107] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1The function specified in one or more boxes.

[0108] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0109] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0110] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A management method based on campus information security, characterized in that, The method includes: S100: Obtain the student's original information image and divide the original information image into several original pixel blocks based on the image block mechanism; S200: Construct a multi-dimensional noise model, classify different levels based on the content and sensitivity of different original pixel blocks, and set corresponding noise type parameters and intensity parameters for different levels of original pixel blocks; extract the corresponding noise data from the multi-dimensional noise model and inject it into the corresponding original pixel blocks according to the corresponding noise type parameters and intensity parameters; S300: Re-fuse the injected original pixel blocks and smooth the boundaries between adjacent pixel blocks to form a new face image, and store the new face image; S400: Perform quality assessment on the new face image to obtain an assessment value. If the assessment value is greater than the preset assessment threshold, store the corresponding new face image. If the assessment value is not greater than the preset assessment threshold, repeat steps S300 and S400 until the assessment value is greater than the preset assessment threshold or the preset maximum number of optimizations is reached. S500: Identify occluded parts of the face and parts with changes in expression, and mark them as noise-free raw pixel blocks; S600: Establish a random face image library, calculate the image matching degree between the original information image and all face images in the random face image library, and select the random face image with the highest image matching degree as the interference image; based on the image block mechanism, split the interference image into several interference pixel blocks; the interference pixel blocks correspond one-to-one with the original pixel blocks; continue to execute step S200; S700: Select the interfering pixel block corresponding to the noiseless original pixel block, re-fuse the interfering pixel block and the injected original pixel block, and smooth the boundary between adjacent pixel blocks to form a new face image; continue to execute step S400. The construction of a multidimensional noise model includes: determining the dimensions and initial parameters of the model, selecting a chaotic system to generate noise data, distributing the generated noise data to each dimension of the model according to the injection rules, and combining the dimensions of the distributed noise data to form a multidimensional noise model.

2. The management method based on campus information security according to claim 1, characterized in that, Step S100 also includes: S110: Detect the surface features of the original information image and obtain the feature value. If the feature value is greater than the preset feature threshold, proceed to step S500; if the feature value is not greater than the preset feature threshold, proceed to step S200.

3. The management method based on campus information security according to claim 2, characterized in that, The surface feature recognition includes facial region recognition, facial key point detection, facial occlusion and expression change classification, and provides corresponding factor values ​​based on the recognition results; the feature value is a comprehensive value obtained by weighted summation of all factor values.

4. The management method based on campus information security according to claim 1, characterized in that, In step S300, the boundaries between adjacent pixel blocks are smoothed, including: smoothing using Gaussian filtering, as detailed below: in, This represents the smoothed pixel value; This represents the pixel value at coordinates (x, y) in the image; i and j are the horizontal and vertical offsets in the Gaussian filter kernel, respectively; the values ​​of i and j range from [-k, k], where k is the radius of the filter. This refers to the Gaussian kernel function, and the specific formula is as follows: Where σ is the standard deviation of the Gaussian kernel, which controls the smoothness of the filter; It represents the square of the distance from the neighboring pixel (i, j) to the center pixel (0, 0).

5. A management method based on campus information security according to claim 1, characterized in that, In step S400, the quality assessment metrics include image sharpness, noise level, and information retention; specifically, peak signal-to-noise ratio (PSNR) is used for evaluation, as follows: in, It is an evaluation value. It is the square of the maximum pixel value of the image. This is the mean squared error, calculated as follows: Where M and N are the height and width of the image, respectively.

6. The management method based on campus information security according to claim 1, characterized in that, Step S700 also includes: S710: After re-fusing the interfering pixel blocks and the injected original pixel blocks, an initial image is formed; the deviation type is determined based on the surface feature analysis of the interfering pixel blocks and the injected original pixel blocks, and the deviation value between each pixel block in the initial image is calculated according to the deviation type; S720: Select the corresponding face image from the random face image library according to the deviation value and mark it as the overlay image. Superimpose the overlay image on the initial image to form a new face image. The deviation between each pixel block in the initial image is calculated using the following formula: in, It is the deviation value. This represents the weighting coefficient for the k-th type of deviation. This represents the deviation value for the k-th deviation type. This indicates the total number of deviation types.

7. A management method based on campus information security according to claim 6, characterized in that, The process involves selecting a corresponding face image from a random face image database based on the deviation value and marking it as the overlay image. Specifically, this includes: combining the calculated deviation values ​​to form a deviation value feature vector; extracting the corresponding feature vectors from the face images in the random face image database to form a random face feature vector; calculating the similarity between the deviation value feature vector and the random face feature vector using a similarity measurement method, and selecting the random face image with the highest similarity as the overlay image.

8. A management method based on campus information security according to claim 1, characterized in that, The method further includes: S800: Acquire several new face images, and identify and extract feature regions with a repetition rate greater than the repetition threshold, and mark them as easily repeated regions; S810: Abstracts the areas prone to repetition, fills them with different character codes, stores the corresponding character codes, and forms a new face image.

9. A management method based on campus information security according to claim 8, characterized in that, The process of abstracting the regions prone to duplication includes: converting the regions to grayscale, extracting features from the processed regions, and mapping the extracted region features onto the character set based on a pre-set character set to generate the corresponding character encoding.

Images