Website picture tampering detection method, computer device, medium and product

Abstract

The invention discloses a website picture tampering detection method, a computer device, a medium and a product. The method comprises the following steps: preprocessing a webpage picture to obtain a target picture; judging whether the target picture contains text information or not, if not, obtaining the picture similarity between the target picture and an illegal picture library, if the picture similarity is greater than a first threshold value, triggering picture tampering early warning, and otherwise, not triggering the picture tampering early warning; if the target picture contains the text information, determining text content; obtaining the semantic similarity between the text content and the sensitive word bank; if the semantic similarity is larger than a second threshold value, picture tampering early warning is triggered; and if the semantic similarity is not greater than a second threshold value, not triggering picture tampering early warning. According to the method, illegal and sensitive contents in the website pictures can be effectively detected, each picture can be comprehensively and accurately detected, and the security and compliance of the website are protected.

Description

Technical Field

[0001] The present disclosure relates to the field of computer technology, and in particular to a website image tampering detection method, computer device, medium, and product. Background Art

[0002] In web page tampering, tampering with images in the website is a common means of attack. The detection methods disclosed in the prior art mainly include: extracting image files from the web page source code, extracting text content from the image, matching the extracted text content with a predefined sensitive word library, identifying images containing illegal or sensitive words, monitoring the website's access behavior, and detecting whether there are abnormal image uploads, downloads, or replacements, for example, a large number of images are replaced in a short period of time or the access frequency is abnormal.

[0003] Although the above methods can detect tampered images to a certain extent, there are still several major problems: 1) For illegal images without text, existing technologies cannot extract effective text content, resulting in the inability to identify these images through sensitive word detection; 2) Even highly accurate image classification models may mistakenly identify normal images as illegal images, especially when the image content is unclear or noisy. For new or unknown illegal images, the model may not be able to accurately identify them, resulting in missed reports; 3) For large websites or a large number of images, the image extraction, content analysis and detection process will be very time-consuming, affecting detection efficiency; 4) Attackers may bypass existing detection mechanisms through technical means (such as image encryption, steganography, etc.), or evade detection through minor modifications to the image.

[0004] Although the detection methods for tampered website images in the existing technology are effective to a certain extent, there are still problems such as the inability to identify illegal images without text, false positives and negatives, low detection efficiency, and adversarial attacks. Summary of the invention

[0005] In view of this, the embodiments of the present disclosure provide a website image tampering detection method, computer device, medium, and product, which can solve the problems of poor detection accuracy, low efficiency, high false alarm rate and missed alarm rate existing in the prior art.

[0006] In a first aspect, the embodiments of the present disclosure provide a method for detecting website image tampering, which specifically includes the following contents:

[0007] Preprocess the acquired web page image to obtain the target image;

[0008] Determine whether the target image contains text information, if not, obtain the similarity between the target image and the image in the illegal image library, and if the image similarity is greater than a first threshold, trigger an image tampering warning;

[0009] If the image similarity is not greater than the first threshold, no image tampering warning is triggered;

[0010] If the target image contains text information, determine the text content;

[0011] Obtaining semantic similarity between the text content and a sensitive word library;

[0012] If the semantic similarity is greater than a second threshold, a picture tampering warning is triggered;

[0013] If the semantic similarity is not greater than the second threshold, the image tampering warning is not triggered.

[0014] Optionally, the preprocessing of the acquired web page image to obtain the target image includes:

[0015] Normalizing the web page image to obtain a sample image of a preset size;

[0016] Data enhancement is performed on the sample image, and resolution processing is performed on the enhanced sample image to obtain a target image with a preset resolution.

[0017] Optionally, obtaining the similarity between the target image and an image in an illegal image library includes:

[0018] Extracting actual picture features of the target picture, wherein the actual picture features include a first color feature vector and a first shape feature vector;

[0019] Extracting sample picture features of each picture in the illegal picture library, wherein the sample picture features include a second color feature vector and a second shape feature vector;

[0020] Analyze the first shape feature vector and the second shape feature vector based on a preset distance formula to obtain a first similarity;

[0021] Convert the first similarity based on a preset conversion formula to obtain a first conversion value;

[0022] Acquire first pixel information based on the first color feature vector;

[0023] Acquire second pixel information based on the second color feature vector;

[0024] Obtaining a second similarity based on the first pixel information and the second pixel information;

[0025] The picture similarity is obtained based on the first conversion value and the second similarity.

[0026] Optionally, obtaining a second similarity based on the first pixel information and the second pixel information includes:

[0027] Based on the first pixel information, first frequency information of all pixels in the target image is obtained; the first frequency information of each pixel includes a first red channel frequency, a first green channel frequency, and a first blue channel frequency;

[0028] Based on all the first frequency information, a first red channel class frequency set, a first green channel class frequency set, and a first blue channel class frequency set are obtained;

[0029] Based on the second pixel information, obtain the second frequency information of all pixels of each picture in the illegal picture library; the second frequency information of each pixel includes a second red channel frequency, a second green channel frequency, and a second blue channel frequency;

[0030] Based on all the second frequency information, a second red channel class frequency set, a second green channel class frequency set, and a second blue channel class frequency set are obtained;

[0031] Based on the first red channel class frequency set and the second red channel class frequency set, obtaining a first channel similarity;

[0032] Based on the first green channel class frequency set and the second green channel class frequency set, obtaining a second channel similarity;

[0033] Based on the first blue channel class frequency set and the second blue channel class frequency set, obtaining a third channel similarity;

[0034] The second similarity is obtained based on the first channel similarity, the second channel similarity, and the third channel similarity.

[0035] Optionally, obtaining the first channel similarity based on the first red channel class frequency set and the second red channel class frequency set includes:

[0036] Based on the first red channel class frequency set and the second red channel class frequency set, obtaining a first channel vector comparison set, wherein the first channel vector comparison information includes red channel comparison information of all pixels;

[0037] Based on the first channel vector comparison set, a first valid set is obtained; the first valid set includes a set of pixels at the same position in the first red channel class frequency set and the second red channel class frequency set;

[0038] Obtaining a first channel similarity based on the first valid set;

[0039] The first channel similarity is a ratio of the first valid set to the total number of red channel pixels.

[0040] Optionally, obtaining the second channel similarity based on the first green channel class frequency set and the second green channel class frequency set includes:

[0041] Based on the first green channel class frequency set and the second green channel class frequency set, a second channel vector comparison set is obtained, wherein the second channel vector comparison information includes green channel comparison information of all pixels;

[0042] Based on the second channel vector comparison set, a second valid set is obtained; the second valid set includes a set of pixels at the same position in the first green channel class frequency set and the second green channel class frequency set;

[0043] Obtaining a second channel similarity based on the second valid set;

[0044] The second channel similarity is a ratio of the second valid set to the total number of green channel pixels.

[0045] Optionally, obtaining a third channel similarity based on the first blue channel class frequency set and the second blue channel class frequency set includes:

[0046] Based on the first blue channel class frequency set and the second blue channel class frequency set, a third channel vector comparison set is obtained, wherein the third channel vector comparison information includes blue channel comparison information of all pixels;

[0047] Based on the third channel vector comparison set, a third valid set is obtained; the third valid set includes a set of pixels at the same position in the first blue channel class frequency set and the second blue channel class frequency set;

[0048] Obtaining a third channel similarity based on the third valid set;

[0049] The third channel similarity is the ratio of the third valid set to the total number of blue channel pixels.

[0050] Optionally, the first similarity is d(A, B): Among them, A i is the first shape eigenvector, B i is the second shape eigenvector;

[0051] The first conversion value is SimShape(x,y):

[0052] The second similarity is SimRGB(x,y):

[0053] Wherein, HR(1,2) is the first channel similarity, HG(1,2) is the second channel similarity, and HB(1,2) is the third channel similarity;

[0054] The image similarity is Sim(x,y):

[0055] Optionally, obtaining the semantic similarity between the text content and a sensitive word library includes:

[0056] Preprocessing the text content to obtain word segmentation information;

[0057] Based on the word segmentation information, obtaining a first text vector;

[0058] Get the known sensitive word vectors of the corresponding images in the sensitive word library;

[0059] Based on the first text vector and the known sensitive word vector, semantic similarity is obtained.

[0060] In a second aspect, the embodiment of the present disclosure further provides a website image tampering detection system, including:

[0061] The acquisition module is used to pre-process the acquired web page image to obtain the target image;

[0062] A text content determination module is used to determine whether the target image contains text information. If not, obtain the similarity between the target image and the image in the illegal image library. If the image similarity is greater than a first threshold, trigger an image tampering warning.

[0063] A first triggering module, configured to obtain target image text when the image similarity is not greater than a first threshold, and trigger a preset detection strategy based on the target image text;

[0064] A second triggering module is used to obtain the target image text when the target image contains text information, and trigger a preset detection strategy based on the target image text;

[0065] The preset detection strategy includes: obtaining the semantic similarity between the target image text and the sensitive word library;

[0066] If the semantic similarity is greater than a second threshold, a picture tampering warning is triggered;

[0067] If the semantic similarity is not greater than the second threshold, the warning of illegal image library is not triggered.

[0068] In a third aspect, the embodiments of the present disclosure further provide a computer device, which adopts the following technical solution:

[0069] The computer device comprises:

[0070] at least one processor; and,

[0071] a memory communicatively connected to the at least one processor; wherein,

[0072] The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute any of the above-mentioned website image tampering detection methods.

[0073] In a fourth aspect, an embodiment of the present disclosure further provides a computer-readable storage medium, which stores computer instructions, and the computer instructions are used to enable a computer to execute any of the above-mentioned website image tampering detection methods.

[0074] In a fifth aspect, an embodiment of the present disclosure further provides a computer program product, including a computer program / instruction, which implements the steps of any of the above methods when executed by a processor.

[0075] The website image tampering detection method disclosed in the present application can improve the quality of images and reduce noise and other interference factors through preprocessing, so that subsequent text detection and similarity comparison are more accurate; by detecting the text content, it can be determined whether the image may contain sensitive information, thereby deciding whether further semantic analysis is needed. If there is no text, the image similarity comparison is directly performed to improve the detection efficiency; through the image similarity comparison, images similar to those in the illegal image library can be quickly identified, the false alarm rate can be reduced, and the detection accuracy can be improved; when text information is included, the text content in the image is extracted to prepare for subsequent semantic analysis to ensure that no possible sensitive information is missed; through semantic similarity analysis, sensitive information in the image can be detected more comprehensively to avoid missed detection due to incomplete matching of the text content with the sensitive word library. Even if the text content does not completely match the sensitive word library, the potential risk can be determined through semantic similarity, effectively avoiding the omission of any possible sensitive information, and achieving comprehensive and accurate detection.

[0076] The above description is only an overview of the technical solution of the present invention. In order to more clearly understand the technical means of the present invention, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present invention more obvious and easy to understand, the following preferred embodiments are specifically cited and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0077] In order to more clearly illustrate the technical solutions of the embodiments of the present disclosure, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present disclosure. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0078] Figure 1 A schematic diagram of the process of detecting website image tampering provided by an embodiment of the present disclosure.

[0079] Figure 2 A flowchart of a method for acquiring a target image provided by an embodiment of the present disclosure.

[0080] Figure 3 A schematic diagram of a flow chart of a method for obtaining image similarity provided in an embodiment of the present disclosure.

[0081] Figure 4 A flowchart of a method for obtaining the similarity between a target image and an illegal image library based on a target twin network model provided in an embodiment of the present disclosure.

[0082] Figure 5 A schematic flow chart of a method for obtaining a second similarity provided in an embodiment of the present disclosure.

[0083] Figure 6 A schematic flow chart of a method for obtaining first channel similarity provided in an embodiment of the present disclosure.

[0084] Figure 7 A schematic diagram of a flow chart of a method for obtaining second channel similarity provided in an embodiment of the present disclosure.

[0086] Figure 8 A schematic diagram of a flow chart of a method for obtaining third channel similarity provided in an embodiment of the present disclosure.

[0087] Fig. 9 A flowchart of a method for obtaining semantic similarity provided in an embodiment of the present disclosure.

[0088] Fig.10 A schematic diagram of the structure of a computer device provided in an embodiment of the present disclosure. DETAILED DESCRIPTION

[0089] The embodiments of the present disclosure are described in detail below with reference to the accompanying drawings.

[0090] It should be clear that the following embodiments of the present disclosure are described by specific specific examples, and those skilled in the art can easily understand other advantages and effects of the present disclosure from the contents disclosed in this specification. Obviously, the described embodiments are only a part of the embodiments of the present disclosure, rather than all of the embodiments. The present disclosure can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present disclosure. It should be noted that the following embodiments and features in the embodiments can be combined with each other in the absence of conflict. Based on the embodiments in the present disclosure, all other embodiments obtained by ordinary technicians in the field without making creative work are within the scope of protection of the present disclosure.

[0091] It should be noted that various aspects of the embodiments within the scope of the appended claims are described below. It should be apparent that the aspects described herein may be embodied in a wide variety of forms, and any specific structure and / or function described herein is merely illustrative. Based on the present disclosure, it should be understood by those skilled in the art that an aspect described herein may be implemented independently of any other aspect, and two or more of these aspects may be combined in various ways. For example, any number of aspects described herein may be used to implement the device and / or practice the method. In addition, other structures and / or functionalities other than one or more of the aspects described herein may be used to implement this device and / or practice this method.

[0092] It should also be noted that the illustrations provided in the following embodiments are only schematic illustrations of the basic concept of the present disclosure. The drawings only show components related to the present disclosure rather than being drawn according to the number, shape and size of components in actual implementation. In actual implementation, the type, quantity and proportion of each component may be changed arbitrarily, and the component layout may also be more complicated.

[0093] Additionally, in the following description, specific details are provided to facilitate a thorough understanding of the examples. However, it will be understood by those skilled in the art that the aspects described may be practiced without these specific details.

[0094] Reference Figure 1 The present application discloses a method for detecting website image tampering, which specifically includes the following steps:

[0095] Preprocess the acquired web page image to obtain the target image;

[0096] Determine whether the target image contains text information. If not, obtain the similarity between the target image and the image in the illegal image library. If the image similarity is greater than a first threshold, trigger an image tampering warning.

[0097] If the image similarity is not greater than the first threshold, the image tampering warning is not triggered.

[0098] If the target image contains text information, determine the text content;

[0099] Obtain the semantic similarity between the text content and the sensitive word library;

[0100] If the semantic similarity is greater than the second threshold, an image tampering warning is triggered;

[0101] If the semantic similarity is not greater than the second threshold, the image tampering warning is not triggered.

[0102] The text information includes one or more characters; the text content refers to the text specifically contained in the target image.

[0103] Among them, web page images can be obtained by crawling; preprocessing can include denoising, grayscale, binarization and other operations; for example, use the OpenCV library to denoise the image to reduce the impact of noise on subsequent analysis; then, convert the image into a grayscale image to reduce the interference of color information on the analysis, and finally perform binarization to convert the image into black and white to facilitate subsequent text detection.

[0104] If the image similarity is greater than the first threshold, it means that the two images are very similar, that is, the image exists in the illegal image library and there is a possibility of tampering. At this time, it is considered that the image has been tampered with, and an early warning is issued. It is manually confirmed whether the early warning is effective. For example, a normal image on a web page is tampered with into an image in an illegal library. We obtain the image on the web page as the target image and perform similarity matching with the image in the illegal library. At this time, if the similarity between the target image and the image in the illegal library is 1 or greater than the set similarity threshold, it indicates that the image on the web page exists in the illegal library or is extremely similar to the image in the illegal library. It can be determined that the originally normal image on the web page has been illegally tampered with, and an early warning is issued.

[0105] If the semantic similarity is greater than the second threshold, it indicates that the text content of the image contains sensitive word information, thereby determining that the image has been tampered with. For example, a picture on a web page is tampered with by adding some sensitive content to the picture. During detection, it is found that the text content of the picture contains sensitive words, then there is a high possibility that the picture has been illegally tampered with.

[0106] The website image tampering detection method disclosed in the present application can improve the quality of images and reduce noise and other interference factors through preprocessing, so that subsequent text detection and similarity comparison are more accurate; by detecting the text content, it can be determined whether the image may contain sensitive information, thereby deciding whether further semantic analysis is needed. If there is no text, the image similarity comparison is directly performed to improve the detection efficiency; through the image similarity comparison, images similar to those in the illegal image library can be quickly identified, the false alarm rate can be reduced, and the detection accuracy can be improved; when text information is included, the text content in the image is extracted to prepare for subsequent semantic analysis to ensure that no possible sensitive information is missed; through semantic similarity analysis, sensitive information in the image can be detected more comprehensively to avoid missed detection due to incomplete matching of the text content with the sensitive word library. Even if the text content does not completely match the sensitive word library, the potential risk can be determined through semantic similarity, effectively avoiding the omission of any possible sensitive information, and achieving comprehensive and accurate detection.

[0107] The method disclosed in the present application, by combining image similarity comparison and semantic analysis, can not only quickly identify illegal images, but also accurately determine the potential sensitive information in the images, thereby improving the accuracy of detection; through a multi-step detection process, it can avoid false alarms that may be caused by a single detection method and reduce the false alarm rate; for images without text content, image similarity comparison can be directly performed to improve detection efficiency; through semantic similarity analysis, it can identify sensitive information even if it does not completely match the sensitive word library, thereby ensuring the comprehensiveness of detection; by setting different thresholds, the sensitivity and strictness of detection can be adjusted according to actual needs, and it has high flexibility; the method disclosed in the present application can detect images embedded with sensitive words and illegal images, which can effectively improve the accuracy of web page image tampering and reduce the underreporting of web page image tampering.

[0108] Reference Figure 2 , the method for obtaining the target image specifically includes the following steps:

[0109] S110, normalizing the web page image to obtain a sample image of a preset size.

[0110] Specifically, normalization can ensure that all web page images have a uniform size, which helps with subsequent image processing and analysis, while reducing noise and distortion in the images and improving image quality; a uniform size can reduce analysis errors caused by different image sizes; and by unifying the image size to a preset size, the computational complexity of image processing can be reduced.

[0111] S120, performing data enhancement on the sample image, and performing resolution processing on the enhanced sample image to obtain a target image with a preset resolution.

[0112] Specifically, data augmentation can increase the diversity of training data by performing random transformations on sample images (such as rotation, scaling, cropping, color change, etc.) to simulate various changes that may occur in images in the real world, thereby improving the ability to adapt to different situations.

[0113] By processing the enhanced sample images into a preset resolution, it can be ensured that all target images have consistent image quality, which is helpful for subsequent similarity calculation and semantic analysis. High-resolution images will bring high computational complexity. By unifying the images to the preset resolution, the use of computing resources can be reduced and the overall detection efficiency can be improved. In addition, the unified resolution helps to reduce the analysis errors caused by resolution differences, especially when performing image similarity comparison and semantic analysis, a fixed resolution can improve detection accuracy.

[0114] The target image acquisition method disclosed in this embodiment standardizes the image size and quality through normalization processing and data enhancement; resolution processing further optimizes the image quality, which can effectively reduce computational complexity and improve detection efficiency and accuracy.

[0115] Reference Figure 3 , the method for obtaining the image similarity specifically includes the following steps:

[0116] A100, building an illegal image library.

[0117] Building an illegal image library provides a standardized reference set for detection. By collecting and annotating a large number of illegal images, it can ensure that detection has a clear target to learn and compare. Furthermore, the illegal image library can be dynamically updated based on new threats and attack methods to ensure that the model can detect the latest illegal images.

[0118] A200, trains the constructed initial network model based on the illegal image library, and uses the trained initial network model as the target twin network model.

[0119] Using illegal image libraries to train the initial network model can quickly and effectively enable the model to learn the characteristics of illegal images. This training method based on a specific data set can significantly improve the detection accuracy of the model.

[0120] A300, obtains the image similarity between the target image and the illegal image library based on the target twin network model.

[0121] Specifically, refer to Figure 4 ,The method for obtaining the similarity between the target image and the image in the illegal image library based on the target twin network model includes the following steps:

[0122] A310, extracting actual picture features of the target picture, where the actual picture features include a first color feature vector and a first shape feature vector.

[0123] A320, extracting sample image features of each image in the illegal image library, where the sample image features include a second color feature vector and a second shape feature vector.

[0124] By extracting actual color features, actual shape features, sample color features, and sample shape features, the features of the image can be comprehensively analyzed from multiple dimensions to increase the accuracy and robustness of detection; the extraction of color features and shape features helps to analyze the image in a refined manner and can better identify subtle differences and features in the image.

[0125] A330: Analyze the first shape feature vector and the second shape feature vector based on a preset distance formula to obtain a first similarity.

[0126] Specifically, the first similarity is d(A,B): Among them, A i is the first shape eigenvector, B i is the second shape eigenvector.

[0127] A340: Convert the first similarity based on a preset conversion formula to obtain a first conversion value.

[0128] Specifically, the first conversion value is SimShape(x,y):

[0129] A350, obtaining first pixel information based on the first color feature vector;

[0130] Acquire second pixel information based on the second color feature vector;

[0131] A second similarity is obtained based on the first pixel information and the second pixel information.

[0132] By analyzing the shape feature vector through a preset distance formula, the similarity between the shapes of two images can be quantified. This method can effectively identify images that are similar in shape.

[0133] By comparing the first pixel information and the second pixel information, a second similarity can be obtained. The comparison of pixel information can capture subtle differences in the images and increase the accuracy of detection.

[0134] A360 obtains the image similarity based on the first conversion value and the second similarity.

[0135] Specifically, the image similarity is Sim(x,y):

[0136] SimRGB(x,y) is the second similarity.

[0137] By combining the first conversion value and the second similarity, the final image similarity can be obtained. This multi-level integration can improve the comprehensiveness and accuracy of similarity calculation; combining the similarities of different features can better identify target images similar to those in the illegal image library and enhance the overall detection capability.

[0138] The method disclosed in this embodiment obtains the similarity between a target image and an image in an illegal image library based on a target twin network model. Through multi-step feature extraction, feature vector acquisition, similarity calculation, similarity conversion and comprehensive similarity, the method for obtaining image similarity can comprehensively and accurately analyze the similarity between the target image and the images in the illegal image library. This method not only improves the accuracy and robustness of detection, but also can effectively identify subtle differences and new illegal images, thereby enhancing the overall security protection capability.

[0139] Reference Figure 5 , the method for obtaining the second similarity comprises the following steps:

[0140] B100, based on the first pixel information, obtaining first frequency information of all pixels; the first frequency information of each pixel includes a first red channel frequency, a first green channel frequency, and a first blue channel frequency.

[0141] In the step, the first frequency information of all pixels in the target image is obtained.

[0142] B200: Based on all the first frequency information, obtain a first red channel class frequency set, a first green channel class frequency set, and a first blue channel class frequency set.

[0143] B300, based on the second pixel information, obtain the second frequency information of all pixels; the second frequency information of each pixel includes the second red channel frequency, the second green channel frequency, and the second blue channel frequency.

[0144] In the step, the first frequency information of all pixels of each picture in the illegal picture library is obtained.

[0145] B400, based on all the second frequency information, obtain a second red channel class frequency set, a second green channel class frequency set, and a second blue channel class frequency set.

[0146] By obtaining the frequency information of the red, green and blue channels for each pixel, the color characteristics of the image can be analyzed in detail. This method can capture subtle differences in color and improve the accuracy of similarity calculation. In addition, the frequency information of the red, green and blue channels provides multi-dimensional color characteristics, which helps to comprehensively compare the color similarity of images.

[0147] By aggregating the frequency information of all pixels, we can obtain the class frequency sets of red, green and blue channels. This aggregation can integrate the scattered pixel information into a more representative feature set, which is convenient for subsequent similarity calculation.

[0148] B500, based on the first red channel class frequency set and the second red channel class frequency set, obtain the first channel similarity.

[0149] B600, obtaining a second channel similarity based on the first green channel class frequency set and the second green channel class frequency set.

[0150] B700, obtaining a third channel similarity based on the first blue channel class frequency set and the second blue channel class frequency set.

[0151] By calculating the similarity of the red, green and blue channels separately, the similarity of each color channel can be analyzed independently, and the similarity differences on different color channels can be captured to improve the accuracy of detection; the similarity calculation of independent channels can refine the color features of the image and help identify subtle differences in color.

[0152] B800, obtaining a second similarity based on the first channel similarity, the second channel similarity, and the third channel similarity.

[0153] Specifically, the second similarity is SimRGB(x,y):

[0154] Among them, HR(1,2) is the first channel similarity, HG(1,2) is the second channel similarity, and HB(1,2) is the third channel similarity.

[0155] By integrating the similarities of the red, green and blue channels, a comprehensive second similarity can be obtained, which can integrate the similarity information of different color channels and improve the comprehensiveness and accuracy of the similarity calculation; by averaging the similarities of the three channels, the impact of different channels on the similarity calculation can be balanced to ensure the fairness and accuracy of the detection results.

[0156] The method for obtaining the second similarity disclosed in this embodiment can comprehensively and accurately analyze the color similarity of images through frequency information extraction, frequency set construction, channel similarity calculation and comprehensive channel similarity. This method can not only improve the accuracy and refinement of detection, but also effectively identify subtle differences in color and new types of illegal images.

[0157] Reference Figure 6 , the method for obtaining the first channel similarity specifically includes:

[0158] B510, based on the first red channel class frequency set and the second red channel class frequency set, obtain a first channel vector comparison set, the first channel vector comparison information includes red channel comparison information of all pixels.

[0159] By constructing the first channel vector comparison set, the data of the first red channel class frequency set and the second red channel class frequency set can be structured to facilitate subsequent analysis and processing; the first channel vector comparison set contains the red channel comparison information of all pixels, which provides a basis for the subsequent effective set extraction and ensures the comprehensiveness and accuracy of subsequent steps.

[0160] B520, based on the first channel vector comparison set, obtain a first valid set; the first valid set includes a set of pixels at the same position in the first red channel class frequency set and the second red channel class frequency set.

[0161] By obtaining the first valid set and retaining only the set of pixels with the same position in the first red channel class frequency set and the second red channel class frequency set, the accuracy of the comparison can be ensured and noise interference can be reduced; the construction of the valid set helps to reduce the amount of data that needs to be processed in subsequent calculations, thereby improving calculation efficiency.

[0162] B530, obtaining a first channel similarity based on the first valid set;

[0163] The first channel similarity is the ratio of the first valid set to the total number of red channel pixels.

[0164] By calculating the ratio of the first valid set to the total number of pixels in the red channel, the similarity of the first channel can be intuitively quantified. This quantification method can help quickly understand the similarity between two images on the red channel. The calculation method of the first channel similarity can truly reflect the similarity between the red channels of the images, making the similarity value interpretable and easy to understand and apply.

[0165] Through the above steps, the method for obtaining the similarity of the first channel can comprehensively and accurately analyze the similarity between the target image and the red channel in the illegal image library. This method not only reduces noise and interference and improves the accuracy of the data, but also makes the detection results easier to interpret and apply through a concise ratio calculation method, thereby enhancing the overall detection capability and practicality.

[0166] Reference Figure 7 , the method for obtaining the second channel similarity includes:

[0167] B610, based on the first green channel class frequency set and the second green channel class frequency set, obtain a second channel vector comparison set, the second channel vector comparison information includes green channel comparison information of all pixels.

[0168] By constructing the second channel vector comparison set, the data of the first green channel class frequency set and the second green channel class frequency set can be structured to facilitate subsequent analysis and processing; the second channel vector comparison set contains the green channel comparison information of all pixels, which provides a basis for the subsequent valid set extraction and ensures the comprehensiveness and accuracy of subsequent steps.

[0169] B620, based on the second channel vector comparison set, obtain a second valid set; the second valid set includes a set of pixel points with the same position in the first green channel class frequency set and the second green channel class frequency set.

[0170] By obtaining the second valid set and retaining only the set of pixels with the same position in the first green channel class frequency set and the second green channel class frequency set, the accuracy of the comparison can be ensured and noise interference can be reduced; the construction of the valid set helps to reduce the amount of data that needs to be processed in subsequent calculations, thereby improving calculation efficiency.

[0171] B630, obtaining a second channel similarity based on the second valid set;

[0172] The second channel similarity is the ratio of the second valid set to the total number of green channel pixels.

[0173] By calculating the ratio of the second effective set to the total number of green channel pixels, the similarity of the second channel can be intuitively quantified; the calculation method of the second channel similarity can truly reflect the similarity between the green channels of the images, making the similarity value interpretable and easy to understand and apply.

[0174] Through the above steps, the method for obtaining the second channel similarity can comprehensively and accurately analyze the similarity between the target image and the green channel in the illegal image library. This method not only reduces noise and interference and improves the accuracy of the data, but also makes the detection results easier to interpret and apply through a concise ratio calculation method, thereby enhancing the overall detection capability and practicality.

[0175] Reference Figure 8 , the method for obtaining the third channel similarity includes:

[0176] B710, based on the first blue channel class frequency set and the second blue channel class frequency set, obtain a third channel vector comparison set, the third channel vector comparison information including blue channel comparison information of all pixels.

[0177] By constructing the third channel vector comparison set, the data of the first blue channel class frequency set and the second blue channel class frequency set can be structured to facilitate subsequent analysis and processing; the third channel vector comparison set contains the blue channel comparison information of all pixels, which provides a basis for the subsequent effective set extraction and ensures the comprehensiveness and accuracy of subsequent steps.

[0178] B720, based on the third channel vector comparison set, obtain a third valid set; the third valid set includes a set of pixels at the same position in the first blue channel class frequency set and the second blue channel class frequency set.

[0179] By obtaining the third valid set, only the set of pixels with the same position in the first blue channel class frequency set and the second blue channel class frequency set is retained, which can ensure the accuracy of the comparison and reduce noise interference; the construction of the valid set helps to reduce the amount of data that needs to be processed in subsequent calculations, from YY+242299P

[0180] And improve computing efficiency.

[0181] B730, obtaining a third channel similarity based on a third valid set;

[0182] The third channel similarity is the ratio of the third valid set to the total number of blue channel pixels.

[0183] By calculating the ratio of the third valid set to the total number of blue channel pixels, the similarity of the third channel can be intuitively quantified. This quantification method can help quickly understand the similarity between two images in the blue channel; the calculation method of the third channel similarity can truly reflect the similarity between the blue channels of the images, making the similarity value interpretable and easy to understand and apply.

[0184] Through the above steps, the method for obtaining the third channel similarity can comprehensively and accurately analyze the similarity between the target image and the blue channel in the illegal image library. This method can not only reduce noise and interference and improve the accuracy of the data, but also make the detection results easier to interpret and apply through a concise ratio calculation method, thereby enhancing the overall detection capability and practicality.

[0185] The method for obtaining the second similarity is described in detail below: calculate the RGB colors of image 1 (i.e., the actual image) and image 2 (i.e., the sample image) respectively, and define three arrays HR (i.e., the frequency of the red channel), HG (i.e., the frequency of the green channel), and HB (i.e., the frequency of the blue channel); assuming that the RGB value of a pixel point P is (4, 231, 129), then the calculation is HR[4]+=1, HG

[231] +=1, HB

[129] +=1. After completing the statistics for each pixel point, we get: HR1(r11, r12, r13, …,r1n) (i.e., the first red channel class frequency set), HR2 (r21, r22, r23, …, r2n) (i.e., the second red channel class frequency set), HG1 (g11, g12, g13, …, g1n) (i.e., the first green channel class frequency set), HG2 (g21, g22, g23, …, g2n) (i.e., the second green channel class frequency set), HB1 (b11, b12, b13, …, b1n) (i.e., the first blue channel class frequency set), HB2 (b21, b22, b23, …, b2n) (i.e., the second blue channel class frequency set).

[0186] Calculate the similarity of the three colors of picture 1 and picture 2 respectively. The calculation method is as follows: compare the corresponding values ​​of vector HR1 (i.e., the first red channel class frequency set) and HR2 (i.e., the second red channel class frequency set) to see if they are equal. If they are equal, the corresponding position is recorded as 1, otherwise it is recorded as 0. The final result is a set such as HR12 (1, 0, 1, 0, ..., 1), HG12 (1, 0, 1, 0, 0, ..., 0), and HB12 (1, 0, 1, 1, ..., 1). Finally, calculate HR The similarity HR12 calculates the final value Sim12 = (HR12[1] + HR12[2] + ... HR[n]) / n (i.e. the first channel similarity). Similarly, the similarity values ​​of HG12 (i.e. the second channel similarity) and HB12 (i.e. the second channel similarity) can be calculated. Finally, the color similarity (i.e. the second similarity) of picture 1 and picture 2 is calculated based on the three color similarities: SimRGB(1,2) = (HR12+HG12+HB12) / 3.

[0187] Reference Fig. 9 , the method for obtaining semantic similarity specifically includes:

[0188] S410, preprocessing the text content to obtain word segmentation information.

[0189] The preprocessing includes word segmentation, stop word removal, etc.; the word segmentation information includes the vector representation of each word.

[0190] Specifically, the mechanical word segmentation method can be used to obtain the word segmentation information, that is, the character string in the text is matched, and if it can be found in the dictionary, the word is segmented out.

[0191] Through preprocessing and word segmentation, the target image text data can be standardized, irrelevant characters and punctuation can be removed, and the accuracy of subsequent processing can be improved; the text data after word segmentation is easier to vectorize, which improves processing efficiency.

[0192] S420: Obtain a first text vector based on the word segmentation information.

[0193] Specifically, based on the word segmentation information, the vector of each word in the corpus can be calculated, and the Skip-gram algorithm can be used for training.

[0194] The goal of the Skip-Gram model is to maximize the probability of the occurrence of context words given the target word, that is, to maximize the following objective function:

[0195] Where T is the total number of words in the text, c is the context size, and w t is the target word, w t+j is the context word, and this formula expresses the probability of predicting the target word through the context word.

[0196] The vector representation of the text (ie, the first text vector) can be obtained by averaging the vector representations of all words in the text (ie, the vectors of all words included in the word segmentation information).

[0197] For a given text, suppose we have a set V = {v1, v2, ..., v n}, where v i Represents the vector of the i-th word in the text. Then, the average vector of all words can be calculated by the following formula:

[0198] Here, n is the number of words in the text.

[0199] Convert the text information after word segmentation into vector form to facilitate mathematical operations and similarity calculations. This numerical representation can capture the semantic information in the text and provide a basis for subsequent similarity analysis; text vectorization can capture the high-dimensional features of the text and improve the accuracy of similarity calculations.

[0200] S430, obtaining a known sensitive word vector corresponding to the image in the sensitive word library.

[0201] By obtaining the known sensitive word vectors of the corresponding images in the sensitive word library, we can use the existing sensitive word data to perform similarity calculations, thereby improving the accuracy and pertinence of the calculations. The known sensitive word vectors provide reliable data support for similarity calculations, making the calculation results more valuable for reference.

[0202] S440, obtaining semantic similarity based on the first text vector and the known sensitive word vector.

[0203] Specifically, the semantic similarity includes: cos(θ)=A·B / (||A||*||B||), where A and B represent the first text vector, the known sensitive word vector and the sensitive word vector, respectively.

[0204] By calculating the similarity between the first text vector and the known sensitive word vector, the semantic relationship between the target image text and the sensitive words in the sensitive word library can be quantified. This quantification method can intuitively reflect the degree of semantic similarity; the calculation method of semantic similarity can efficiently detect whether the target image text contains sensitive content, thereby improving the efficiency and accuracy of detection.

[0205] Through text preprocessing and word segmentation, text vectorization, obtaining known sensitive word vectors and semantic similarity calculation, the semantic similarity acquisition method can comprehensively and accurately analyze the semantic similarity between the target image text and the sensitive words in the sensitive word library. This method not only improves the accuracy and efficiency of detection, but also quantifies the semantic relationship, making the detection results easier to interpret and apply, thereby enhancing the overall security protection capabilities.

[0206] In this application, after the warning is triggered, it can be manually confirmed whether it has actually been tampered with.

[0207] In the present application, for target images without text content, by calculating the similarity between the target image and the images in the illegal image library, if the similarity exceeds a first threshold, an image tampering warning is triggered. This step can identify those images that have no text but are still illegal; this method not only relies on the matching of sensitive word libraries, but also introduces image similarity comparison and semantic similarity analysis. Through multi-level detection, the false alarm rate (such as normal images are misjudged as illegal) and the missed alarm rate (such as new or unknown illegal images are not identified) can be reduced; by setting reasonable thresholds (first threshold and second threshold), false alarms and missed alarms can be minimized while ensuring detection accuracy.

[0208] This method first determines whether the image contains text content. If not, it directly performs image similarity comparison, avoiding unnecessary text extraction and analysis, thereby improving detection efficiency; if the target image contains text content, this method will first obtain the image text, and then perform semantic similarity analysis. This selective processing method can reduce the computational burden of comprehensive analysis of all images and improve overall detection efficiency; by combining image similarity comparison and semantic similarity analysis, this method can detect images that bypass existing detection mechanisms through technical means (such as image encryption, steganography, etc.), even if the attacker evades detection through minor modifications, image similarity comparison can still detect changes in the image; among them, the illegal image library can be updated regularly to add new and unknown illegal image samples, thereby improving the robustness and accuracy of detection. The website image tampering detection method disclosed in this application effectively solves the problems of being unable to identify illegal images without text, false positives and false negatives, low detection efficiency, and adversarial attacks in the prior art by combining image similarity comparison and semantic similarity analysis. This method not only improves the accuracy and efficiency of detection, but also can cope with more complex attack methods and enhance the overall security protection capability.

[0209] In a second aspect, the present application discloses a website image tampering detection system, comprising:

[0210] The acquisition module is used to pre-process the acquired web page image to obtain the target image;

[0211] A text content determination module is used to determine whether the target image contains text information. If not, the target image and the image similarity in the illegal image library are obtained. If the image similarity is greater than a first threshold, an image tampering warning is triggered.

[0212] A first trigger module, configured to obtain target image text when the image similarity is not greater than a first threshold, and trigger a preset detection strategy based on the target image text;

[0213] A second trigger module is used to obtain the target image text when the target image contains text information, and trigger a preset detection strategy based on the target image text;

[0214] The preset detection strategy includes: obtaining the semantic similarity between the target image text and the sensitive word library;

[0215] If the semantic similarity is greater than the second threshold, an image tampering warning is triggered;

[0216] If the semantic similarity is not greater than the second threshold, the warning of illegal image library is not triggered.

[0217] It should be noted that the scheme in the website image tampering detection method disclosed in the first aspect of the present application is also applicable to the website image tampering detection system disclosed in the second aspect of the present application, so it will not be repeated here.

[0218] The computer device according to the embodiment of the present disclosure includes a memory and a processor. The memory is used to store non-temporary computer-readable instructions. Specifically, the memory may include one or more computer program products, and the computer program product may include various forms of computer-readable storage media, such as volatile memory and / or non-volatile memory. The volatile memory may, for example, include a random access memory (RAM) and / or a cache memory (cache), etc. The non-volatile memory may, for example, include a read-only memory (ROM), a hard disk, a flash memory, etc.

[0219] The processor may be a central processing unit (CPU) or other forms of processing units with data processing capabilities and / or instruction execution capabilities, and may control other components in the computer device to perform desired functions. In one embodiment of the present disclosure, the processor is used to run the computer-readable instructions stored in the memory, so that the computer device executes all or part of the steps of the website image tampering detection method of each embodiment of the present disclosure.

[0220] Those skilled in the art should be able to understand that in order to solve the technical problem of how to obtain a good user experience, the present embodiment may also include well-known structures such as a communication bus and an interface, and these well-known structures should also be included in the protection scope of the present disclosure.

[0221] like Fig.10 A schematic diagram of the structure of a computer device provided in an embodiment of the present disclosure is shown, which is a schematic diagram of the structure of a computer device suitable for implementing the embodiment of the present disclosure. Fig.10 The computer device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present disclosure.

[0222] like Fig.10 As shown, the computer device may include a processor (e.g., a central processing unit, a graphics processing unit, etc.), which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) or a program loaded from a storage device into a random access memory (RAM). In the RAM, various programs and data required for the operation of the computer device are also stored. The processor, ROM, and RAM are connected to each other via a bus. An input / output (I / O) interface is also connected to the bus.

[0223] Typically, the following devices can be connected to the I / O interface: input devices such as sensors or visual information acquisition devices; output devices such as display screens; storage devices such as tapes, hard disks, etc.; and communication devices. The communication device can allow the computer device to communicate with other devices (such as edge computing devices) wirelessly or by wire to exchange data. Fig.10 A computer device having various devices is shown, but it should be understood that it is not required to implement or possess all of the devices shown. More or fewer devices may be implemented or possessed instead.

[0224] In particular, according to an embodiment of the present disclosure, the process described above with reference to the flowchart can be implemented as a computer software program. For example, an embodiment of the present disclosure includes a computer program product, which includes a computer program carried on a non-transitory computer-readable medium, and the computer program contains program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network through a communication device, or installed from a storage device, or installed from a ROM. When the computer program is executed by a processor, all or part of the steps of the website image tampering detection method of the embodiment of the present disclosure are executed.

[0225] For detailed description of this embodiment, reference may be made to the corresponding descriptions in the aforementioned embodiments, which will not be repeated here.

[0226] According to the computer-readable storage medium of the embodiment of the present disclosure, non-transitory computer-readable instructions are stored thereon. When the non-transitory computer-readable instructions are executed by the processor, all or part of the steps of the website image tampering detection method of each embodiment of the present disclosure are executed.

[0227] The above-mentioned computer-readable storage media include, but are not limited to: optical storage media (e.g., CD-ROM and DVD), magneto-optical storage media (e.g., MO), magnetic storage media (e.g., magnetic tape or mobile hard disk), media with built-in rewritable non-volatile memory (e.g., memory card) and media with built-in ROM (e.g., ROM box).

[0228] For detailed description of this embodiment, reference may be made to the corresponding descriptions in the aforementioned embodiments, which will not be repeated here.

[0229] The basic principles of the present disclosure are described above in conjunction with specific embodiments. However, it should be noted that the advantages, strengths, effects, etc. mentioned in the present disclosure are only examples and not limitations, and it cannot be considered that these advantages, strengths, effects, etc. are required by each embodiment of the present disclosure. In addition, the specific details disclosed above are only for the purpose of illustration and ease of understanding, and are not limitations. The above details do not limit the present disclosure to the necessity of adopting the above specific details to be implemented.

[0230] In the present disclosure, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply that there is any such actual relationship or order between these entities or operations. The block diagrams of the devices, devices, equipment, and systems involved in the present disclosure are only illustrative examples and are not intended to require or imply that they must be connected, arranged, and configured in the manner shown in the block diagram. As will be appreciated by those skilled in the art, these devices, devices, equipment, and systems can be connected, arranged, and configured in any manner. Words such as "including", "comprising", "having", etc. are open words, referring to "including but not limited to", and can be used interchangeably with them. The words "or" and "and" used here refer to the words "and / or" and can be used interchangeably with them, unless the context clearly indicates otherwise. The words "such as" used here refer to the phrase "such as but not limited to", and can be used interchangeably with them.

[0231] Additionally, as used herein, "or" used in a list of items beginning with "at least one" indicates a separate list, so that, for example, a list of "at least one of A, B, or C" means A or B or C, or AB or AC or BC, or ABC (i.e., A and B and C). Furthermore, the word "exemplary" does not mean that the example described is preferred or better than other examples.

[0232] It should also be noted that in the system and method of the present disclosure, each component or each step can be decomposed and / or recombined. Such decomposition and / or recombination should be regarded as equivalent solutions of the present disclosure.

[0233] Various changes, substitutions, and modifications of the techniques described herein may be made without departing from the teachings defined by the appended claims. Furthermore, the scope of the claims of the present disclosure is not limited to the specific aspects of the processes, machines, manufactures, compositions of events, means, methods, and actions described above. Currently existing or later to be developed processes, machines, manufactures, compositions of events, means, methods, or actions that perform substantially the same functions or achieve substantially the same results as the corresponding aspects described herein may be utilized. Thus, the appended claims include such processes, machines, manufactures, compositions of events, means, methods, or actions within their scope.

[0234] The above description of the disclosed aspects is provided to enable any person skilled in the art to make or use the present disclosure. Various modifications to these aspects will be readily apparent to those skilled in the art, and the general principles defined herein may be applied to other aspects without departing from the scope of the present disclosure. Therefore, the present disclosure is not intended to be limited to the aspects shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

[0235] The above description has been given for the purpose of illustration and description. In addition, this description is not intended to limit the embodiments of the present disclosure to the forms disclosed herein. Although multiple example aspects and embodiments have been discussed above, those skilled in the art will recognize certain variations, modifications, changes, additions and sub-combinations thereof.

Claims

1. A website image tampering detection method, characterized in that: include: Preprocess the acquired web page image to obtain the target image; Determine whether the target image contains text information. If not, obtain the similarity between the target image and the image in the illegal image library. If the image similarity is greater than a first threshold, trigger an image tampering warning. If the image similarity is not greater than the first threshold, no image tampering warning is triggered; If the target image contains text information, determine the text content; Obtaining semantic similarity between the text content and a sensitive word library; If the semantic similarity is greater than a second threshold, a picture tampering warning is triggered; If the semantic similarity is not greater than the second threshold, the image tampering warning is not triggered.

2. The website image tampering detection method according to claim 1, characterized in that: The step of preprocessing the acquired web page image to obtain the target image includes: Normalizing the web page image to obtain a sample image of a preset size; Data enhancement is performed on the sample image, and resolution processing is performed on the enhanced sample image to obtain a target image with a preset resolution.

3. The website image tampering detection method according to claim 1, characterized in that: The obtaining of the similarity between the target image and the image in the illegal image library includes: Extracting actual picture features of the target picture, wherein the actual picture features include a first color feature vector and a first shape feature vector; Extracting sample picture features of each picture in the illegal picture library, wherein the sample picture features include a second color feature vector and a second shape feature vector; Analyze the first shape feature vector and the second shape feature vector based on a preset distance formula to obtain a first similarity; Convert the first similarity based on a preset conversion formula to obtain a first conversion value; Acquire first pixel information based on the first color feature vector; Acquire second pixel information based on the second color feature vector; Obtaining a second similarity based on the first pixel information and the second pixel information; The picture similarity is obtained based on the first conversion value and the second similarity.

4. The website image tampering detection method according to claim 3 is characterized in that: The obtaining a second similarity based on the first pixel information and the second pixel information includes: Based on the first pixel information, first frequency information of all pixels in the target image is obtained; the first frequency information of each pixel includes a first red channel frequency, a first green channel frequency, and a first blue channel frequency; Based on all the first frequency information, a first red channel class frequency set, a first green channel class frequency set, and a first blue channel class frequency set are obtained; Based on the second pixel information, obtain the second frequency information of all pixels of each picture in the illegal picture library; the second frequency information of each pixel includes a second red channel frequency, a second green channel frequency, and a second blue channel frequency; Based on all the second frequency information, a second red channel class frequency set, a second green channel class frequency set, and a second blue channel class frequency set are obtained; Based on the first red channel class frequency set and the second red channel class frequency set, obtaining a first channel similarity; Based on the first green channel class frequency set and the second green channel class frequency set, obtaining a second channel similarity; Based on the first blue channel class frequency set and the second blue channel class frequency set, obtaining a third channel similarity; The second similarity is obtained based on the first channel similarity, the second channel similarity, and the third channel similarity.

5. The website image tampering detection method according to claim 4, characterized in that: The obtaining of the first channel similarity based on the first red channel class frequency set and the second red channel class frequency set comprises: Based on the first red channel class frequency set and the second red channel class frequency set, obtaining a first channel vector comparison set, wherein the first channel vector comparison information includes red channel comparison information of all pixels; Based on the first channel vector comparison set, a first valid set is obtained; the first valid set includes a set of pixels at the same position in the first red channel class frequency set and the second red channel class frequency set; Obtaining a first channel similarity based on the first valid set; The first channel similarity is a ratio of the first valid set to the total number of red channel pixels.

6. The website image tampering detection method according to claim 5, characterized in that: The obtaining of the second channel similarity based on the first green channel class frequency set and the second green channel class frequency set includes: Based on the first green channel class frequency set and the second green channel class frequency set, a second channel vector comparison set is obtained, wherein the second channel vector comparison information includes green channel comparison information of all pixels; Based on the second channel vector comparison set, a second valid set is obtained; the second valid set includes a set of pixels at the same position in the first green channel class frequency set and the second green channel class frequency set; Obtaining a second channel similarity based on the second valid set; The second channel similarity is a ratio of the second valid set to the total number of green channel pixels.

7. The website image tampering detection method according to claim 6, characterized in that: The obtaining a third channel similarity based on the first blue channel class frequency set and the second blue channel class frequency set includes: Based on the first blue channel class frequency set and the second blue channel class frequency set, a third channel vector comparison set is obtained, wherein the third channel vector comparison information includes blue channel comparison information of all pixels; Based on the third channel vector comparison set, a third valid set is obtained; the third valid set includes a set of pixels at the same position in the first blue channel class frequency set and the second blue channel class frequency set; Obtaining a third channel similarity based on the third valid set; The third channel similarity is the ratio of the third valid set to the total number of blue channel pixels.

8. The website image tampering detection method according to claim 7, characterized in that: The first similarity is d(A,B): Among them, A i is the first shape eigenvector, B i is the second shape eigenvector; The first conversion value is SimShape(x,y): The second similarity is SimRGB(x,y): Wherein, HR(1,2) is the first channel similarity, HG(1,2) is the second channel similarity, and HB(1,2) is the third channel similarity; The image similarity is Sim(x,y):

9. The website image tampering detection method according to claim 1, characterized in that: The obtaining of the semantic similarity between the text content and the sensitive word library includes: Preprocessing the text content to obtain word segmentation information; Based on the word segmentation information, obtaining a first text vector; Get the known sensitive word vectors of the corresponding images in the sensitive word library; Based on the first text vector and the known sensitive word vector, semantic similarity is obtained.

10. A computer device, characterized in that: The computer device comprises: at least one processor; and, a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the website image tampering detection method described in any one of claims 1-9.

11. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a computer to execute the website image tampering detection method described in any one of claims 1-9.

12. A computer program product comprising computer instructions, characterized in that: When the computer instructions are executed by a processor, the steps of the method according to any one of claims 1 to 9 are implemented.

Images