File privacy risk assessment method and device and electronic equipment

Abstract

The invention discloses a file privacy risk assessment method and apparatus, and an electronic device. The method comprises the steps of obtaining basic information of various types of target files, association information between the target files, and interaction information between the target files; determining a type weight corresponding to each type of target file according to the basic information; determining a type privacy risk value corresponding to each type of target file according to the basic information and the associated information; determining an interaction weight and an interaction privacy risk value corresponding to each type of target file according to the interaction information; and according to the type weight, the type privacy risk value, the interaction weight and the interaction privacy risk value, determining an overall privacy risk value of each type of target file.

Description

Technical Field

[0001] The present application belongs to the field of artificial intelligence, and specifically relates to a method, device and electronic device for assessing file privacy risks. Background Art

[0002] In related technologies, file privacy leakage detection is mainly based on analyzing the content of files, such as through natural language processing, image processing, and audio processing, to identify sensitive information in files. However, these methods ignore the multi-dimensional associations between different files, making it impossible to accurately assess the privacy leakage risks of files in complex scenarios with multiple files. Summary of the invention

[0003] The embodiments of the present application provide a method, device and electronic device for assessing file privacy risks, which can solve the problem in the related art that the detection of file privacy leakage is based on the content of the file but ignores the multi-dimensional associations between different files, resulting in the inability to accurately assess the privacy leakage risk.

[0004] In a first aspect, an embodiment of the present application provides a method for assessing file privacy risks, the method comprising: obtaining basic information of target files of various types, association information between the target files, and interaction information between the target files; determining type weights corresponding to the target files of various types based on the basic information; determining type privacy risk values ​​corresponding to the target files of various types based on the basic information and the association information; and determining interaction weights and interaction privacy risk values ​​corresponding to the target files of various types based on the interaction information; determining the overall privacy risk values ​​of the target files of various types based on the type weights, the type privacy risk values, the interaction weights, and the interaction privacy risk values.

[0005] In a second aspect, an embodiment of the present application provides a device for evaluating file privacy risks, which includes: an acquisition module for acquiring basic information of target files of various types, association information between the target files, and interaction information between the target files; a determination module for determining type weights corresponding to the target files of various types based on the basic information; determining type privacy risk values ​​corresponding to the target files of various types based on the basic information and the association information; and determining interaction weights and interaction privacy risk values ​​corresponding to the target files of various types based on the interaction information; an evaluation module for determining the overall privacy risk value of the target files of various types based on the type weights, the type privacy risk values, the interaction weights, and the interaction privacy risk values.

[0006] In a third aspect, an embodiment of the present application provides an electronic device, which includes a processor, a memory, and a program or instruction stored in the memory and executable on the processor, wherein the program or instruction, when executed by the processor, implements the steps of the method described in the first aspect.

[0007] In a fourth aspect, an embodiment of the present application provides a readable storage medium, on which a program or instruction is stored, and when the program or instruction is executed by a processor, the steps of the method described in the first aspect are implemented.

[0008] In a fifth aspect, an embodiment of the present application provides a chip, comprising a processor and a communication interface, wherein the communication interface is coupled to the processor, and the processor is used to run a program or instruction to implement the steps of the method described in the first aspect.

[0009] In a sixth aspect, an embodiment of the present application provides a computer program product, comprising a computer program stored on a non-transitory computer-readable storage medium, wherein the computer program comprises a program or instructions, and when the program or instructions are executed, the steps of the method described in the first aspect are implemented.

[0010] In an embodiment of the present application, by obtaining basic information of target files of various types, association information between the target files, and interaction information between the target files, the type weight corresponding to the target files of various types is determined according to the basic information; the type privacy risk value corresponding to the target files of various types is determined according to the basic information and the association information; and the interaction weight and interaction privacy risk value corresponding to the target files of various types are determined according to the interaction information, thereby determining the overall privacy risk value of the target files of various types according to the type weight, the type privacy risk value, the interaction weight, and the interaction privacy risk value. The evaluation of the privacy risk of files in the embodiment of the present application introduces the association information and interaction information of the files of various types, which can more comprehensively understand the relationship between the files of various types, so as to more accurately evaluate the privacy risk of the files, so as to provide corresponding privacy protection suggestions based on the obtained privacy risk value. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] Figure 1 It is a flowchart of a method for assessing file privacy risk provided by an embodiment of the present application; Figure 2 It is a flowchart of another method for assessing file privacy risk provided by an embodiment of the present application; Figure 3 It is a structural schematic diagram of a file privacy risk assessment device provided in an embodiment of the present application; Figure 4It is a structural schematic diagram of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0012] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0013] The terms "first", "second", etc. in the specification and claims of the present application are used to distinguish similar objects, and are not used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable under appropriate circumstances, so that the embodiments of the present application can be implemented in an order other than those illustrated or described here, and the objects distinguished by "first", "second", etc. are generally of one type, and the number of objects is not limited. For example, the first object can be one or more. In addition, "and / or" in the specification and claims represents at least one of the connected objects, and the character " / " generally indicates that the objects associated with each other are in an "or" relationship.

[0014] The following, in conjunction with the accompanying drawings, describes in detail the file privacy risk assessment method, device and electronic device provided by the embodiments of the present application through specific embodiments and their application scenarios.

[0015] Figure 1 The flowchart of a method for assessing file privacy risk provided by an embodiment of the present application is shown, and the method can be executed by an electronic device. Figure 1 , the method may include the following steps.

[0016] Step 102: acquiring basic information of target files of various types, association information between the target files, and interaction information between the target files.

[0017] Among them, each type of target file is each file that will be evaluated for privacy risk value. The basic information of each type of target file may include the text and symbols of text type files, the bitmap information header and pixel data of image type files, the metadata and bit rate of audio type files, etc. The association information between the target files may include the association between the text content of text type files and images, audio, etc. The interactive information between the target files may include the conversion between different types of files, and the embedding of a certain type of file into another type of file as part of its content. For example, a text type file may be embedded with images, audio or video, etc.

[0018] The above information can be obtained from multiple data sources, which may include: data stored in the user's cloud disk, data from social media accounts, application usage record data, data obtained by device sensors, etc. The scope of acquisition of the above information depends on the acquisition configuration and user permission to ensure the privacy and security of the acquired data. The types of target files may include: text, such as user's text messages, social media posts, emails, etc.; images, such as pictures uploaded by users, image content in text messages or social media; audio, such as recordings, voice messages, audio files, etc.; video, such as video content uploaded by users; device sensor data, such as geographic location, user heart rate, and number of steps.

[0019] Step 104: determine the type weight corresponding to the target files of each type according to the basic information; determine the type privacy risk value corresponding to the target files of each type according to the basic information and the associated information; and determine the interaction weight and interaction privacy risk value corresponding to the target files of each type according to the interaction information.

[0020] Among them, the type weight represents the relative importance of different types of target files to the overall privacy risk value. For example, the basic information includes the privacy sensitivity of each type of target file. The privacy sensitivity of the image is higher than the privacy sensitivity of the text, so the corresponding image type weight is greater than the text type weight. The type privacy risk value represents the privacy risk value of different types of target files, which can be calculated by algorithms or models, such as machine learning models, rule engines or other computing technologies. The interaction weight represents the relative importance of different types of target files to the overall privacy risk value when there is interaction between them. The interaction privacy risk value represents the privacy risk value when there is interaction between different types of target files, which can be calculated by algorithms or models, such as machine learning models, rule engines or other computing technologies.

[0021] Step 106: Determine the overall privacy risk value of the target files of each type according to the type weight, the type privacy risk value, the interaction weight, and the interaction privacy risk value.

[0022] For example, based on the type weight, the type privacy risk value, the interaction weight and the interaction privacy risk value, the overall privacy risk value of the target files of each type can be determined by C_score = ∑(W_type·Type_score) + ∑(W_interaction · Interaction_score), wherein C_score represents the overall privacy risk value, W_type represents the type weight, Type_score represents the type privacy risk value, W_interaction represents the interaction weight, and Interaction_score represents the interaction privacy risk value.

[0023] In an embodiment of the present application, by obtaining basic information of target files of various types, association information between the target files, and interaction information between the target files, the type weight corresponding to the target files of various types is determined according to the basic information; the type privacy risk value corresponding to the target files of various types is determined according to the basic information and the association information; and the interaction weight and interaction privacy risk value corresponding to the target files of various types are determined according to the interaction information, thereby determining the overall privacy risk value of the target files of various types according to the type weight, the type privacy risk value, the interaction weight, and the interaction privacy risk value. The evaluation of the privacy risk of files in the embodiment of the present application introduces the association information and interaction information of various types of files, so that it not only takes into account the characteristics of the target files of various types themselves, but also takes into account the impact of the association and interaction between the target files on the privacy risk, so that the privacy risk of the files can be evaluated more accurately, so as to provide corresponding privacy protection suggestions based on the obtained privacy risk value.

[0024] In one example, based on the basic information and the pre-trained deep neural network model, the type weights corresponding to the target files of each type are determined. Based on the basic information and the associated information and the pre-trained graph neural network model, the type privacy risk values ​​corresponding to the target files of each type are determined. For example, for text files, W_type = 0.7, Type_score = 0.8; for image files, W_type = 0.5, Type_score = 0.7; for audio files, W_type = 0.4, Type_score = 0.6. Based on the interaction information and the pre-trained multi-head self-attention model, the interaction weights corresponding to the target files of each type combination are determined. Based on the interaction information and the pre-trained generative adversarial network model, the interaction information is input into the pre-trained generative adversarial network model. For example, W_interaction = 0.6, Interaction_score = 0.7. Thus, the overall privacy risk value C_score = (0.7 · 0.8 + 0.5 · 0.7 + 0.4·0.6) +(0.6 · 0.7) = 0.56 + 0.42 = 0.98 can be calculated. According to the preset privacy risk threshold or range, the overall privacy risk value is divided into different risk levels, such as low, medium, and high. Then, based on the privacy risk level, privacy protection suggestions corresponding to the privacy risk level are provided.

[0025] In one implementation, the step 104 of determining the type privacy risk value corresponding to each type of target file according to the basic information and the associated information may include: inputting the basic information and the associated information into a pre-trained graph neural network model, and outputting the type privacy risk value corresponding to each type of target file for each type of target file; Among them, each node of the graph neural network model represents each type of target file, and each edge of the graph neural network model represents the association relationship between each type of target files.

[0026] Optionally, before inputting the basic information and the associated information into a pre-trained graph neural network model, the basic information and the associated information are preprocessed to ensure that the basic information and the associated information are suitable for processing by the graph neural network model.

[0027] In an embodiment of the present application, basic information may include the type of target file, type distribution, and content of the target file, etc., and associated information may include whether different types of target files have the same, similar, or related content, etc. The basic information and associated information are input into a pre-trained graph neural network model, and different types of target files become nodes in the graph neural network model. Target files with associated relationships are connected by edges, thereby completing the data input of the graph neural network model. Processing the input data through each layer of the graph neural network model helps to analyze and understand the association relationship between the target files, improve the accuracy of determining the type privacy risk value, and output the type privacy risk value corresponding to each type of target file for subsequent evaluation of the overall privacy risk value.

[0028] In one example, the user's cloud disk data includes text files (T1, T2, T3), image files (I1, I2, I3), and audio files (A1, A2, A3). Determining the type privacy risk value of each type of file may include the following steps: S1. Obtain the following information from the user's cloud disk data: (1) Basic information: The user has 3 text files, 3 image files, and 3 audio files. (2) Association information between files: T1 is associated with I1. S2. Preprocess the obtained basic information and association information. S3. Input the preprocessed basic information and association information into a pre-trained graph neural network model, such as a multi-level graph neural network model, and output the type privacy risk value corresponding to each type of file after processing at each layer. For example, Type_score(T) = 0.7, Type_score(I) = 0.6, Type_score(A) = 0.4. Type_score reflects the type privacy risk assessment value of each file. In this example, text files have the highest privacy risk, followed by image files, and the lowest is audio files.

[0029] In one implementation, the above-mentioned step 104 determines the type weights corresponding to the target files of each type based on the basic information, and may include: inputting the basic information into a pre-trained deep neural network model, and outputting the type weights corresponding to the target files of each type for each type of the target files.

[0030] Optionally, before inputting the basic information into the pre-trained deep neural network model, the basic information is preprocessed to ensure that the basic information and associated information are suitable for processing by the deep neural network model.

[0031] In an embodiment of the present application, the basic information may include the type of target file, the content of the target file, and the user's interest level, etc. The basic information is input into a pre-trained deep neural network model, and the input basic information is processed through each layer of the deep neural network model to output the type weights corresponding to each type of target file for subsequent evaluation of the overall privacy risk value.

[0032] In one example, the user's cloud disk data includes text files, image files, and audio files. Determining the type weight of each type of file may include the following steps.

[0033] S1. Obtain basic information from the user's cloud disk data: the user has 10 text files, 5 image files and 3 audio files.

[0034] S2. Preprocess the acquired basic information, such as performing word segmentation on text files, extracting image features on image files, performing spectrum processing on audio files, etc.

[0035] S3. Input the pre-processed basic information into the pre-trained deep neural network model, and after processing at each layer, output the type weights corresponding to each type of file. For example, W_type(T) = 0.7, W_type(I) = 0.5, W_type(A) = 0.4. W_type reflects the relative contribution of each file type to the overall privacy risk.

[0036] In this example, text files contribute the most to the overall privacy risk, followed by image files, and the lowest is audio files.

[0037] In one implementation, the above step 104 determines the interaction weights corresponding to the target files of each type according to the interaction information, which may include: inputting the interaction information into a pre-trained multi-head self-attention model, and outputting the interaction weights corresponding to the target files of each type combination for the target files of each type combination; Wherein, each head of the multi-head self-attention model represents each type of target file, and the type combination includes at least two types of target files. For example, the type combination can be: text-audio, text-image, text-audio-video, etc.

[0038] Optionally, before inputting the interaction information into a pre-trained multi-head self-attention model, the interaction information is preprocessed to ensure that the interaction information is suitable for processing by the multi-head self-attention model.

[0039] In an embodiment of the present application, the interaction information may include the user's associated behaviors on at least two types of target files based on the association relationship between the target files. For example, the user browsed a text file containing "A" content, and browsed an image file containing "A" content. The interaction information is input into a pre-trained multi-head self-attention model, where each head of the model represents a target file of each type. The input basic information is processed by each layer of the multi-head self-attention model, and the interaction weight corresponding to each type of target file combination is output for subsequent evaluation of the overall privacy risk value.

[0040] In one example, the user's cloud disk data includes text files (T1, T2), image files (I), and audio files (A). The type combinations may be: text file (T)-image file (I), text file (T)-audio file (A), and image file (I)-audio file (A). Determining the interaction weights of each type of file may include the following steps.

[0041] S1. Obtain interaction information from the user's cloud disk data: the user recently uploaded 1 text file (T), 1 image file (I), and 1 audio file (A); the user searched and browsed text files (T1) and image files (I) in the past week; the user modified image files (I) and audio files (A) in the past week; the user deleted text files (T2) and audio files (A) in the past week.

[0042] S2. Preprocess the acquired interaction information, for example, convert the interaction information into a vector form.

[0043] S3. Input the pre-processed interaction information into the pre-trained multi-head self-attention model, and after processing at each layer, output the interaction weights corresponding to each type of combination of files. For example, W_interaction(T, I) = 0.6, W_interaction(T, A) = 0.3, W_interaction(I, A) = 0.4. Among them, W_interaction reflects the relative contribution of the interaction between different types of files to the overall privacy risk.

[0044] In this example, the interaction between text files and image files contributes the most to the overall privacy risk, followed by image files and audio files, and the least to text files and audio files.

[0045] In one implementation, the above step 104 determines the interactive privacy risk values ​​corresponding to the target files of each type according to the interactive information, which may include: inputting the interactive information into a pre-trained generative adversarial network model, and outputting the interactive privacy risk values ​​corresponding to the target files of each type combination for the target files of each type combination; The type combination includes at least two types of target files.

[0046] Optionally, before inputting the interaction information into a pre-trained multi-head self-attention model, the interaction information is preprocessed to ensure that the interaction information is suitable for processing by the multi-head self-attention model.

[0047] In the embodiment of the present application, the interaction information may include the interaction frequency between different types of target files, the associated behavior of the user on at least two types of target files based on the associated relationship between the target files, etc. By inputting the interaction information into the pre-trained generative adversarial network model, the generative adversarial network model identifies the associated relationship between each target file, processes the interaction information through each layer, and outputs the interaction privacy risk value corresponding to each type combination of target files, so as to be used for the subsequent evaluation of the overall privacy risk value.

[0048] In one example, the user's cloud disk data includes text files (T1, T2, T3), image files (I1, I2, I3), and audio files (A1, A2, A3). Among them, the type combination can be: text file (T)-image file (I), text file (T)-image file (I)-audio file (A). Determining the interactive privacy risk value of each type of file may include the following steps.

[0049] S1. Obtain interaction information from the user's cloud disk data: There is interaction between T1 and I1, and there is interaction between T2 and A2.

[0050] S2. Preprocess the acquired interaction information, for example, convert the interaction information into a vector form.

[0051] S3. Input the pre-processed interaction information into the pre-trained generative adversarial network model, and after processing at each layer, output the interaction privacy risk value Interaction_score corresponding to each type combination of files. Interaction_score reflects the interaction privacy risk assessment value of each type combination.

[0052] In one implementation, the above step 102 of obtaining the association information between the target files may include the following steps.

[0053] Step 1021: input the target files of each type into a neural network model corresponding to the type of the target files, and extract target features corresponding to the target files of each type.

[0054] Among them, for text files, a recurrent neural network model can be used for feature extraction, and the extracted text features can include keywords, sentiment analysis, themes, etc. of the text files. For image, video, and audio files, a convolutional neural network model can be used for feature extraction, and the extracted image features can include objects, colors, textures, etc. in the image files; the extracted audio features can include voice content, voice emotions, spectrum, etc.; the extracted video features can include key frames, object traces in video files, object behaviors, etc.; and geographic location features can be extracted from data acquired by device sensors. By processing different types of target files through a neural network model corresponding to the type of target file, the accuracy and efficiency of feature extraction can be improved.

[0055] Step 1022, calculating the similarity between the target features.

[0056] The association relationship between different types of target files is identified by analyzing the similarity between each target feature, thereby obtaining the association information.

[0057] In an optional implementation, by mapping each target feature into a shared vector space, the target features corresponding to target files of different types can be used on the same measurement scale to achieve cross-modal similarity calculation, thereby measuring the correlation between target files of different types and obtaining correlation information.

[0058] Step 1023: Determine association information between the target files of each type according to the similarity.

[0059] In an embodiment of the present application, the target files of various types are input into a neural network model corresponding to the type of the target files, and the target features corresponding to the target files of various types are extracted, so that the target files of different types are converted into target features that can be compared in the same dimension, and the similarities between the target features are calculated, so that the association information between the target files of various types can be determined based on the similarities. For example, if the similarity between the features corresponding to a picture file and the features corresponding to a text file is greater than a preset threshold, it can be indicated that the picture file and the text file have a certain association relationship, and the association information between the picture file and the text file can be determined.

[0060] In one implementation, before calculating the similarity between the target features in step 1022, the method further includes: inputting the target features into a multimodal autoencoder for multimodal embedding and mapping them into a shared vector space.

[0061] The shared vector space includes feature vectors corresponding to the respective target features.

[0062] In an embodiment of the present application, by inputting the various target features into a multimodal autoencoder for multimodal embedding and mapping them into a shared vector space, target files of various types can be compared, analyzed and processed under the same measurement scale, for example, calculating the similarity between target features.

[0063] In one example, analyzing a photo and a text document and identifying the relationship between them may include the following steps.

[0064] S1. Feature extraction: Extract visual features from images, such as color and shape in photos. Extract text features from text, such as words and sentences in documents.

[0065] S2, map visual features and text features into a shared vector space.

[0066] S3. Calculate the similarity between visual features and text features to determine the association between photos and text documents, that is, positively identify whether there is an association between two types of files.

[0067] S4. Find common occurrence patterns. For example, when a word appears frequently in a text document, what pictures might be related to it, that is, reverse search for files with associated relationships.

[0068] In one implementation, the above step 102 of obtaining the interaction information between the target files may include the following steps.

[0069] Step 1024, obtaining user behavior data; the behavior data is used to characterize the user's behavior on the target file.

[0070] The user's behavior data may include the user's search, browsing, modification, and other behaviors on different types of target files, as well as the frequency of each behavior.

[0071] Step 1025: cluster the behavior data based on a cluster analysis algorithm to obtain a plurality of clusters.

[0072] The obtained multiple clusters have different characteristics, and the behavior data in the same cluster have the same, similar or related characteristics.

[0073] Step 1026: Search the association relationship in each cluster based on the association rule mining algorithm to determine the interaction information between the target files.

[0074] Among them, association rule mining algorithms such as Apriori algorithm or FP-growth algorithm are used to discover frequent associations between different types of files. This will reveal the potential relationships between different types of files, such as certain types of text content are associated with specific image themes and there are interactions.

[0075] In the embodiment of the present application, the acquired user behavior data is clustered based on a cluster analysis algorithm to obtain multiple clusters, so that the behavior data in each cluster has the same, similar or related characteristics, and then the association relationship in each cluster is searched based on an association rule mining algorithm, so that the association relationship of the target file can be identified from a large amount of user behavior data to determine the interactive information between the target files. And the efficiency of identifying the interactive information between the target files is improved by first cluster analysis and then association rule mining.

[0076] In one implementation, after determining the interaction information between the target files in the above step 1026, the method further includes the following steps.

[0077] Step 1027: Establish a personalized knowledge base for the user.

[0078] The personalized knowledge base is established for the user, and the personalized knowledge base stores information representing the user's interest level and preferred content.

[0079] Step 1028: Store the user's behavior data, target files associated with the behavior data, and interaction information between the target files into the personalized knowledge base.

[0080] The user's behavior data may include the user's search, browsing, modification and other behaviors on different types of target files, as well as the frequency of each behavior. For example, if the target file associated with the behavior data is a text file T1 and an image file I1, the text file T1 and the image file I1 are the target files associated with the behavior data. The interaction information between the target files (text file T1 and image file I1) is determined according to the above steps 1024 to 1026.

[0081] In an embodiment of the present application, a personalized knowledge base is established for the user, and the obtained behavior data of the user, the target file associated with the behavior data, and the interaction information between the target files are stored in the personalized knowledge base. This information characterizes the personalized needs of the user and the association between different types of target files, so that when evaluating the file privacy risk for the user, the interaction information related to the user can be quickly obtained from the personalized knowledge base to determine the interaction weight and interaction privacy risk value corresponding to the target file. By introducing personalized weight calculation and personalized knowledge base establishment, a targeted privacy risk assessment method is constructed for each user, thereby realizing personalized privacy risk assessment of target files and improving user experience.

[0082] In one implementation, the above method may further include periodically acquiring the latest behavior data of the user; and updating the personalized knowledge base based on the latest behavior data and an incremental learning algorithm.

[0083] In an embodiment of the present application, based on the changing characteristics of user needs, interests, preferences, etc., by regularly acquiring the user's latest behavior data, the user's needs, interests, preferences, etc. in the current cycle can be captured, and then based on the latest behavior data and the incremental learning algorithm, the personalized knowledge base is updated, that is, according to the user's changing behavior data, various types of information in the personalized knowledge base corresponding to the user are dynamically updated to the latest information, so as to ensure that the personalized knowledge base can timely reflect the user's latest needs, interests, preferences, etc., so that the latest information can always be used when evaluating the file privacy risk value, thereby improving the real-time and accuracy of evaluating the file privacy risk value, and helping to timely discover new privacy risks.

[0084] In an optional implementation, the above-mentioned graph neural network model, deep neural network model, multi-head self-attention model, generative adversarial network model and neural network model corresponding to the type of target file can be trained by introducing adversarial sample data to make each model more robust. On this basis, data processed by new technologies is continuously obtained and used as training data for training models. Each model is updated regularly to improve the accuracy of each model in processing target files.

[0085] Figure 2 A flowchart of another method for evaluating the privacy risk value of a file provided in an embodiment of the present application is shown in FIG. Figure 2 , the method may include the following steps.

[0086] Step 201 : obtaining relevant data of the document to be evaluated and establishing a personalized knowledge base for the user.

[0087] Step 202: Evaluate the privacy risk value of the file to be evaluated in multiple dimensions and calculate the personalized weight.

[0088] Among them, the multi-dimensional information of the files to be evaluated is mined, different types of files to be evaluated, such as text files, image files, audio files, etc., are mapped to the shared vector space, and the associations between different types of files to be evaluated are analyzed. According to the basic information of the files to be evaluated (such as file type, content), association information, and interactive information obtained from the personalized knowledge base, the type privacy risk value and interactive privacy risk value of the files to be evaluated are determined, and the personalized weights, namely the type weight and interactive weight, are calculated.

[0089] Step 203: Evaluate the overall privacy risk value and provide personalized reminders.

[0090] Among them, for each type of file to be evaluated, the overall privacy risk value is determined according to the type privacy risk value and the interactive privacy risk value, as well as the type weight and the interaction weight, that is, it can be determined by the multi-dimensional combined risk calculation formula C_score= ∑(W_type · Type_score) + ∑(W_interaction · Interaction_score), where C_score represents the overall privacy risk value, W_type represents the type weight, Type_score represents the type privacy risk value, W_interaction represents the interaction weight, and Interaction_score represents the interaction privacy risk value. Then, according to the privacy risk range to which the overall privacy risk value belongs, different privacy risk levels are determined to remind users of the risk of privacy leakage of the files to be evaluated and provide corresponding privacy protection suggestions.

[0091] In the embodiment of the present application, a multi-dimensional combined risk calculation formula is introduced, which comprehensively considers the type privacy risk value and interactive privacy risk value of different types of files, and can comprehensively evaluate the privacy risk value of files. In addition, a personalized knowledge base is established for users, which can better meet the personalized needs, interests and preferences of users, and the personalized knowledge base is dynamically updated according to the latest behavior data of users, which can cope with the scenario of changes in user behavior, further improving the accuracy of privacy risk value assessment.

[0092] Optionally, the file privacy risk value assessment method provided in the embodiment of the present application can be applied not only to individual users, but also to enterprises, institutions of various types and fields, etc. that have privacy risk assessment and early warning needs, and can be personalized and expanded for enterprises or institutions to meet the personalized needs of various types of users.

[0093] Through the embodiments of the present application, the privacy risks of different types of target files themselves and the privacy risks of the associations between the target files are comprehensively considered, and the privacy risks of the target files can be comprehensively evaluated from the perspective of multi-dimensional content combination, which more accurately reflects the real and complex application scenarios with privacy risks and improves the accuracy of privacy risk assessment. Based on personalized weight calculation and personalized knowledge base establishment, a targeted privacy risk assessment method is constructed for each user, so that a more accurate privacy risk assessment method can be provided according to the needs of different users. Compared with static privacy risk analysis, by regularly obtaining the latest user behavior data to dynamically adjust the various models used to assess privacy risks, personalized knowledge bases, and privacy risk handling strategies, the accuracy of privacy risk assessment and adaptability to changing data environments are improved.

[0094] It should be noted that the file privacy risk value assessment method provided in the embodiment of the present application can be executed by a file privacy risk value assessment device, or a control module in the file privacy risk value assessment device for executing the file privacy risk value assessment method. In the embodiment of the present application, the file privacy risk value assessment device provided in the embodiment of the present application is taken as an example to illustrate the file privacy risk value assessment device.

[0095] Figure 3 A schematic diagram of the structure of a device for evaluating the privacy risk value of a file provided in an embodiment of the present application is shown. Figure 3 As shown, the device 300 includes: an acquisition module 31 , a determination module 32 and an evaluation module 33 .

[0096] Among them, the acquisition module 31 is used to obtain the basic information of each type of target file, the association information between the target files and the interaction information between the target files; the determination module 32 is used to determine the type weight corresponding to the target file of each type according to the basic information; determine the type privacy risk value corresponding to the target file of each type according to the basic information and the association information; and determine the interaction weight and interaction privacy risk value corresponding to the target file of each type according to the interaction information; the evaluation module 33 is used to determine the overall privacy risk value of the target file of each type according to the type weight, the type privacy risk value, the interaction weight and the interaction privacy risk value.

[0097] In one implementation, the above-mentioned determination module 32 can be used to input the basic information and the associated information into a pre-trained graph neural network model, and output the type privacy risk value corresponding to the target file of each type; wherein each node of the graph neural network model represents the target file of each type, and each edge of the graph neural network model represents the association relationship between the target files of each type.

[0098] In one implementation, the above-mentioned determination module 32 can be used to input the basic information into a pre-trained deep neural network model, and output the type weight corresponding to the target file of each type.

[0099] In one implementation, the above-mentioned determination module 32 can be used to input the interaction information into a pre-trained multi-head self-attention model, and for each type combination of target files, output the interaction weights corresponding to the target files of each type combination; wherein each head of the multi-head self-attention model represents each type of target file; and the type combination includes at least two types of target files.

[0100] In one implementation, the above-mentioned determination module 32 can be used to input the interaction information into a pre-trained generative adversarial network model, and for each type combination of target files, output the interaction privacy risk value corresponding to the target files of each type combination; wherein the type combination includes at least two types of target files.

[0101] In one implementation, the above-mentioned acquisition module 31 can be used to input the target files of each type into a neural network model corresponding to the type of the target files, extract the target features corresponding to the target files of each type; calculate the similarity between the target features; and determine the association information between the target files of each type based on the similarity.

[0102] In one implementation, the acquisition module 31 can also be used to input each target feature into a multimodal autoencoder for multimodal embedding and map it into a shared vector space; wherein the shared vector space includes feature vectors corresponding to each target feature.

[0103] In one implementation, the above-mentioned acquisition module 31 can be used to obtain user behavior data; the behavior data is used to characterize the user's behavior on the target file; the behavior data is clustered based on a clustering analysis algorithm to obtain multiple clusters; based on an association rule mining algorithm, the association relationships in each cluster are searched separately to determine the interaction information between the target files.

[0104] In one implementation, the above-mentioned device 300 may also include a storage module for establishing a personalized knowledge base of the user; storing the user's behavior data, target files associated with the behavior data, and interaction information between the target files in the personalized knowledge base.

[0105] In one implementation, the storage module may also be used to periodically obtain the latest behavior data of the user; and update the personalized knowledge base based on the latest behavior data and an incremental learning algorithm.

[0106] The file privacy risk assessment device in the embodiment of the present application can be a device, or a component, integrated circuit, or chip in a terminal. The device can be a mobile electronic device, and for example, the mobile electronic device can be a mobile phone, a tablet computer, a laptop computer, a PDA, an in-vehicle electronic device, a wearable device, an ultra-mobile personal computer (UMPC), a netbook, or a personal digital assistant (PDA), etc. The non-mobile electronic device can be a server, a network attached storage (NAS), a personal computer (PC), a television (TV), a teller machine or a self-service machine, etc., and the embodiment of the present application does not make specific limitations.

[0107] The file privacy risk assessment device in the embodiment of the present application may be a device having an operating system. The operating system may be an Android operating system, an iOS operating system, or other possible operating systems, which are not specifically limited in the embodiment of the present application.

[0108] The file privacy risk assessment device provided in the embodiment of the present application can achieve Figure 1 and Figure 2 To avoid repetition, the various processes implemented in the method embodiment are not described here.

[0109] The embodiment of the present application further provides an electronic device, which is used to execute the above-mentioned file privacy risk assessment method. Figure 4A schematic diagram of the structure of an electronic device for implementing various embodiments of the present application. The electronic device may have relatively large differences due to different configurations or performances, and may include a processor (processor) 401, a communication interface (Communications Interface) 402, a memory (memory) 403 and a communication bus 404, wherein the processor 401, the communication interface 402, and the memory 403 communicate with each other through the communication bus 404. The processor 401 can call a computer program stored in the memory 403 and can be run on the processor 401 to execute the various steps of the above-mentioned file privacy risk assessment method embodiment, and can achieve the same technical effect, so it will not be repeated here to avoid repetition.

[0110] The above electronic device structure does not constitute a limitation on the electronic device. The electronic device may include more or fewer components than shown in the figure, or combine certain components, or arrange the components differently. For example, the input unit may include a graphics processing unit (GPU) and a microphone, and the display unit may be configured with a display panel in the form of a liquid crystal display, an organic light-emitting diode, etc. The user input unit includes a touch panel and at least one of other input devices. The touch panel is also called a touch screen. Other input devices may include, but are not limited to, a physical keyboard, function keys (such as volume control keys, switch keys, etc.), a trackball, a mouse, and a joystick, which will not be repeated here.

[0111] The memory can be used to store software programs and various data. The memory may mainly include a first storage area for storing programs or instructions and a second storage area for storing data, wherein the first storage area may store an operating system, an application program or instructions required for at least one function (such as a sound playback function, an image playback function, etc.), etc. In addition, the memory may include a volatile memory or a non-volatile memory, or the memory may include both volatile and non-volatile memories. Among them, the non-volatile memory may be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), or a flash memory. Volatile memory can be random access memory (RAM), static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDRSDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous link dynamic random access memory (SLDRAM) and direct memory bus random access memory (DRRAM).

[0112] The processor may include one or more processing units; optionally, the processor integrates an application processor and a modem processor, wherein the application processor mainly processes operations related to the operating system, user interface, and application programs, and the modem processor mainly processes wireless communication signals, such as a baseband processor. It is understandable that the modem processor may not be integrated into the processor.

[0113] An embodiment of the present application also provides a readable storage medium, on which a program or instruction is stored. When the program or instruction is executed by a processor, each process of the above-mentioned file privacy risk assessment method embodiment is implemented, and the same technical effect can be achieved. To avoid repetition, it will not be repeated here.

[0114] The processor is a processor in the electronic device described in the above embodiment. The readable storage medium includes a computer readable storage medium, such as a computer read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0115] An embodiment of the present application further provides a chip, which includes a processor and a communication interface, wherein the communication interface is coupled to the processor, and the processor is used to run programs or instructions to implement the various processes of the above-mentioned file privacy risk assessment method embodiment, and can achieve the same technical effect. To avoid repetition, it will not be repeated here.

[0116] It should be understood that the chip mentioned in the embodiments of the present application can also be called a system-level chip, a system chip, a chip system or a system-on-chip chip, etc.

[0117] An embodiment of the present application also provides a computer program product, which includes a computer program stored on a non-transitory computer-readable storage medium, and the computer program includes a program or instructions. When the program or instructions are executed, the various processes of the above-mentioned file privacy risk assessment method embodiment are implemented, and the same technical effect can be achieved. To avoid repetition, they are not repeated here.

[0118] It should be noted that, in this article, the terms "comprise", "include" or any other variant thereof are intended to cover non-exclusive inclusion, so that the process, method, article or device including a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise one..." do not exclude the presence of other identical elements in the process, method, article or device including the element. In addition, it should be pointed out that the scope of the methods and devices in the embodiments of the present application is not limited to performing functions in the order shown or discussed, and may also include performing functions in a substantially simultaneous manner or in reverse order according to the functions involved, for example, the described method may be performed in an order different from that described, and various steps may also be added, omitted, or combined. In addition, the features described with reference to certain examples may be combined in other examples.

[0119] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, a magnetic disk, or an optical disk), and includes a number of instructions for a terminal (which can be a mobile phone, a computer, a server, an air conditioner, or a network device, etc.) to execute the methods described in each embodiment of the present application.

[0120] The embodiments of the present application are described above in conjunction with the accompanying drawings, but the present application is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the guidance of the present application, ordinary technicians in this field can also make many forms without departing from the purpose of the present application and the scope of protection of the claims, all of which are within the protection of the present application.

Claims

1. A method for assessing file privacy risk, characterized in that: include: Acquire basic information of target files of various types, association information between the target files, and interaction information between the target files; Determine the type weight corresponding to each type of target file according to the basic information; Determine the type privacy risk value corresponding to each type of target file according to the basic information and the associated information; and determining the interaction weights and interaction privacy risk values ​​corresponding to the target files of each type according to the interaction information; The overall privacy risk value of the target files of each type is determined according to the type weight, the type privacy risk value, the interaction weight, and the interaction privacy risk value.

2. The method according to claim 1, characterized in that The determining, according to the basic information and the associated information, the type privacy risk value corresponding to the target file of each type includes: Inputting the basic information and the associated information into a pre-trained graph neural network model, and outputting the type privacy risk value corresponding to each type of target file for each type of target file; Among them, each node of the graph neural network model represents each type of target file, and each edge of the graph neural network model represents the association relationship between each type of target files.

3. The method according to claim 1, characterized in that The determining, according to the basic information, the type weights corresponding to the target files of each type includes: The basic information is input into a pre-trained deep neural network model, and for each type of target file, the type weight corresponding to each type of target file is output.

4. The method according to claim 1, characterized in that: The determining, according to the interaction information, the interaction weights corresponding to the target files of each type includes: Input the interaction information into a pre-trained multi-head self-attention model, and output the interaction weight corresponding to each type combination of target files for each type combination of target files; Among them, each head of the multi-head self-attention model represents each type of target file; the type combination includes at least two types of target files.

5. The method according to claim 1, characterized in that The determining, according to the interaction information, the interaction privacy risk values ​​corresponding to the target files of each type includes: Inputting the interaction information into a pre-trained generative adversarial network model, and outputting, for each type combination of target files, an interaction privacy risk value corresponding to each type combination of target files; The type combination includes at least two types of target files.

6. The method according to claim 1, characterized in that The obtaining of association information between the target files includes: Inputting the target files of each type into a neural network model corresponding to the type of the target files, and extracting target features corresponding to the target files of each type; Calculating the similarity between the target features; According to the similarities, association information between the target files of each type is determined.

7. The method according to claim 6, characterized in that Before calculating the similarity between the target features, the method further includes: Inputting each target feature into a multimodal autoencoder for multimodal embedding and mapping it into a shared vector space; The shared vector space includes feature vectors corresponding to the respective target features.

8. The method according to claim 1, characterized in that The obtaining the interaction information between the target files includes: Acquire user behavior data; the behavior data is used to characterize the user's behavior on the target file; Clustering the behavior data based on a cluster analysis algorithm to obtain a plurality of clusters; Based on the association rule mining algorithm, the association relationship in each cluster is searched respectively to determine the interaction information between the target files.

9. The method according to claim 8, characterized in that After determining the interaction information between the target files, the method further includes: Establishing a personalized knowledge base for the user; The user's behavior data, target files associated with the behavior data, and interaction information between the target files are stored in the personalized knowledge base.

10. The method according to claim 9, characterized in that The method further comprises: Regularly obtain the latest behavior data of the user; Based on the latest behavior data and the incremental learning algorithm, the personalized knowledge base is updated.

11. A device for assessing file privacy risk, characterized in that: include: An acquisition module, used to acquire basic information of target files of various types, association information between the target files, and interaction information between the target files; A determination module, used to determine the type weights corresponding to the target files of each type according to the basic information; Determine the type privacy risk value corresponding to each type of target file according to the basic information and the associated information; and determining the interaction weights and interaction privacy risk values ​​corresponding to the target files of each type according to the interaction information; An evaluation module is used to determine the overall privacy risk value of the target files of each type according to the type weight, the type privacy risk value, the interaction weight and the interaction privacy risk value.

12. An electronic device, characterized in that: The electronic device comprises a processor, a memory, and a program or instruction stored in the memory and executable on the processor, wherein the program or instruction implements the steps of the method according to any one of claims 1 to 10 when executed by the processor.

13. A readable storage medium, characterized in that: The readable storage medium stores a program or instruction, and when the program or instruction is executed by a processor, the steps of the method according to any one of claims 1 to 10 are implemented.

14. A computer program product, characterized in that The computer program product comprises a computer program stored on a non-transitory computer-readable storage medium, wherein the computer program comprises a program or instructions, and when the program or instructions are executed, the steps of the method according to any one of claims 1 to 10 are implemented.

Images