Privacy protection method and system for social network environment information sharing

Abstract

The invention discloses a privacy protection method and system for social network environment information sharing, and the method comprises the steps: building a cascading scale prediction model of privacy information at different time points, and obtaining the propagation scale of the privacy information in time; evaluating the influence of the privacy information receiver of the privacy subject on the privacy subject; the privacy risk is evaluated from two aspects of an objective faced risk and a subjective expected risk; determining a privacy utility based on the privacy revenue and the privacy loss; the privacy utility maximization problem in the online social network is concluded into a constrained Markov decision process to analyze decision earnings at different time points, and a proper strategy is formulated to realize balance. By researching the dynamic characteristics of privacy information propagation, a privacy information propagation scale prediction result is used for privacy subject influence and privacy information sharing risk assessment, and finally a utility maximized privacy protection strategy is made through a Markov decision process to realize privacy information sharing in an online social network.

Description

Technical Field

[0001] The present invention relates to the technical field of data privacy protection, and in particular to a privacy protection method and system for information sharing in a social network environment. Background Art

[0002] Social networks have become a mainstream service for tens of millions of Internet users. Social networks can be used for many types of activities (such as careers, making friends, media sharing, location services, etc.). They have a common feature that users can share information and interact with others. In addition, users in social networks usually fill in their personal information on their homepages, such as gender, age, date of birth, location, etc. Using these social networks brings many benefits to users, such as influencing others, improving reputation, enjoying brand discounts, and establishing connections with a large community. However, one of the disadvantages of using social networks is privacy leakage and its potential consequences. Privacy protection technology can provide users with reliable privacy protection, and has therefore become a key research area for current researchers.

[0003] In the past, privacy-preserving information sharing methods in online social networks can be divided into the following three categories: 1) Privacy-preserving information sharing methods based on access control: This category mainly controls access rights through information attributes, user roles, dynamic access, and access control lists to achieve privacy protection. 2) Privacy-preserving information sharing methods based on differential privacy: This category mainly protects user privacy by adding noise to information content or graph structure while ensuring that the information is still available. 3) Privacy-preserving information sharing methods based on anonymity technology: This category mainly uses anonymity technology as a key method for privacy protection, aiming to prevent personal data leakage and identification by obscuring user identity information.

[0004] A series of existing solutions provide theoretical basis and methodological support for the research topic carried out by this invention. However, the traditional access control mechanism is usually static and difficult to adapt to the dynamically changing user roles, number of users and user relationships in social networks. In addition, the existing privacy protection settings and tools are often too complicated for ordinary users to operate. Users often use the default settings when setting privacy, which results in the privacy protection settings being unable to meet the user's personalized privacy needs, thereby affecting the privacy protection effect. Summary of the invention

[0005] In view of the deficiencies in the prior art, the present invention provides a privacy protection method and system for information sharing in a social network environment. The method studies the dynamic characteristics of private information propagation, uses the prediction results of the scale of private information propagation to assess the influence of privacy subjects and the risk of private information sharing, and finally formulates a privacy protection strategy with maximum utility through a Markov decision process to realize private information sharing in online social networks.

[0006] The present invention discloses a privacy protection method for sharing social network environment information, comprising:

[0007] Step 1: Establish a cascade scale prediction model for private information at different time points to obtain the propagation scale of private information over time;

[0008] Step 2: Based on the sensitivity of the privacy attributes contained in the private information and the propagation scale obtained in step 1, evaluate the influence of the privacy subject's private information receiver on the privacy subject;

[0009] Step 3: Evaluate privacy risks from the perspectives of objective risks and subjective expected risks;

[0010] Step 4: Determine the privacy gain based on the influence obtained in step 2, determine the privacy loss based on the privacy risk obtained in step 3, and determine the privacy utility based on the privacy gain and privacy loss;

[0011] Step 5: The privacy utility maximization problem in online social networks is summarized into a constrained Markov decision process to analyze the decision benefits at different time points and formulate appropriate strategies to achieve balance. In the Markov decision process, the reward function of one step is determined by the current state of the system and the selected action. The privacy utility obtained in step 4 is selected as the reward function in the Markov decision process.

[0012] As a further improvement of the present invention, the step 1 specifically comprises:

[0013] Step 11: Based on the graph convolutional network, the contextualized structural patterns in the information cascade graph and the implicit relationships of users in the social network are captured and modeled to obtain the node feature matrix H GCN ;

[0014] Step 12: Use the self-attention mechanism in Transformer to model the temporal dependency in information propagation and obtain the output result H Trans ;

[0015] Step 13: Use Bayesian neural network to model the changes and uncertainties in information propagation and information growth, and obtain the output result H BNN ;

[0016] Step 14: First, capture the structural information through the graph convolutional network, then capture the time series dependency through the LSTM, and finally make predictions through the multi-layer perceptron to obtain the propagation scale of the private information over time.

[0017] As a further improvement of the present invention,

[0018] In step 11, H GCN The calculation formula is:

[0019]

[0020] In the formula, H GCN represents the node feature matrix obtained by GCN, represents the adjacency matrix after adding self-loops, yes The degree matrix of (l) represents the node feature representation of the lth layer, W (l) is a layer-specific trainable weight matrix, and σ is an activation function that introduces nonlinearity;

[0021] In step 12, H Trans The calculation formula is:

[0022]

[0023] Where Q = W Q H GCN ,K=W K H GCN ,V=W V H GCN are the query, key, and value matrices derived from the input sequence, respectively, and W Q ,W K ,W V are trainable weight matrices for query, key, and value, respectively; d k is the dimension of the key;

[0024] In step 13, H BNN The calculation formula is:

[0025]

[0026] In the formula, H BNN represents the output of the Bayesian neural network, H Trans Represents the output of Transformer, The given data The posterior distribution of the weights, p(y|H Trans ,W) is the given input H Trans and the likelihood of the output of weight W;

[0027] In step 14, the private information I S Prediction results of the spread scale at time t The calculation formula is:

[0028]

[0029] In the formula, Indicates private information IS The propagation scale prediction result at time t, H GCN 、h LSTM and h BNN They represent the outputs of GCN, LSTM, and BNN respectively, concat represents the concatenation operation, and MLP is used for the final prediction.

[0030] As a further improvement of the present invention, in step 2, the influence The calculation formula is:

[0031]

[0032] Where δ is the natural attenuation rate of influence, For privacy information S The scale of spread in online social networks, It is private information S The privacy attributes contained in is the sensitivity of , α and β are weight parameters.

[0033] As a further improvement of the present invention, in step 3, the privacy risk R(I S ,t) is:

[0034] R(I S ,t)=r(I S ,t)+N(I S ,t) (6)

[0035] In the formula, r(I S ,t) represents different time t, I S The objective risks faced; N(I S ,t) represents different time t, I S Risk of subjective expectations;

[0036]

[0037] In the formula, It is private information S The privacy attributes contained in sensitivity, It is private information S the scale of future spread in the current environment;

[0038]

[0039] In the formula, μ is the membership function, Indicates the intention to conceal, r(I S ,t) represents objective risk.

[0040] As a further improvement of the present invention, in step 4, the calculation formula of the privacy utility U(S) is:

[0041]

[0042] In the formula, is the privacy loss. At different times, when the privacy information of the privacy subject S is propagated in OSNs, σ S =1, otherwise σ S =0.

[0043] As a further improvement of the present invention, the step 5 specifically comprises:

[0044] Step 51: In the Markov decision process, the reward function is determined by the current state of the system and the selected action. The utility function U(S) is selected as the reward function in the Markov decision process. The utility function considers the social benefits and privacy losses when sharing and not sharing information, and changes with the choice of the privacy subject S at different decision moments. S and change;

[0045] Step 52: Relative value iteration is used to solve the optimal strategy. First, the Lagrangian method is used to transform the CMDP problem into an unconstrained MDP problem. The utility function U(S) and the risk function R(I) are transformed by introducing the Lagrangian multiplier β. S ,t) to form a Lagrangian reward function l(s,t) to solve the optimal strategy; the above Lagrangian reward function takes into account the balance between privacy utility and risk, and then solves the optimal strategy through the relative value iteration method;

[0046] Step 53: When the CMDP problem is converted into an unconstrained MDP problem, the reward function L(s, t) is used as the objective function for optimization; at this time, the constraints are R(I S ,t) Determine that there is a relationship between the Lagrange multiplier β and the maximum risk; Use the Q-learning algorithm to determine the maximum risk R that the privacy subject can bear max and the corresponding Lagrange multiplier β * , and through β * Control the balance between privacy risk and utility, thereby optimizing the reward function under unconstrained MDP problems and ultimately solving the optimal strategy;

[0047] Step 54: To obtain the optimal strategy, firstly perform Lagrange multiplier β * Perturb and get β - =β * -Δβ and β + =β * +Δβ; then the perturbed β -and β + Substitute the reward function L(S,t) and solve the corresponding two pure strategies δ by relative value iteration method - and δ + , and combined with the utility function U(S) and the risk function R(I S ,t) respectively calculate the average risk R - and R + ; Since the optimal strategy is a mixed strategy of two pure strategies, the optimal strategy can be obtained through a comprehensive balance of utility and risk at each decision moment:

[0048]

[0049] In the formula,

[0050] The present invention also discloses a privacy protection system for sharing social network environment information, including:

[0051] A private information propagation scale prediction module, used to implement step 1 of the above privacy protection method;

[0052] The influence measurement and update module is used to implement step 2 of the above privacy protection method;

[0053] A privacy risk measurement module, used to implement step 3 of the above privacy protection method;

[0054] A privacy utility measurement module, used to implement step 4 of the above privacy protection method;

[0055] The privacy policy decision module is used to implement step 5 of the above privacy protection method.

[0056] Compared with the prior art, the present invention has the following beneficial effects:

[0057] The present invention effectively formulates a privacy protection strategy that conforms to the dynamic propagation of private information by introducing the private information propagation characteristics in online social networks.

[0058] The present invention evaluates the influence of privacy subjects, update methods and privacy sharing risks through the scale of privacy information dissemination, and effectively evaluates the privacy utility that meets the personalized needs of users.

[0059] The present invention maximizes the privacy utility obtained by sharing the benefits and losses of private information by constraining the Markov decision process, thereby formulating a privacy protection strategy that meets the personalized needs of users and achieving the goal of balancing privacy protection and privacy sharing. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] Figure 1 It is a schematic diagram of a privacy protection system for information sharing in a social network environment disclosed by the present invention. DETAILED DESCRIPTION

[0061] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0062] The present invention provides a privacy protection method and system for information sharing in a social network environment, which is a privacy protection information sharing scheme for studying the dynamic propagation of private information in an online social network to promote privacy protection during information sharing; in the privacy protection information sharing scheme based on the propagation of private information, the privacy subject will evaluate different moments in the future and implement the strategy with the greatest utility to the privacy subject; in addition, in order to strike a balance between privacy protection and information sharing, the present invention transforms the problem into a Markov chain decision process by applying reinforcement learning technology to analyze the decision benefits at different time points, so as to formulate appropriate strategies to achieve a balance.

[0063] The present invention is further described in detail below in conjunction with the accompanying drawings:

[0064] The present invention provides a privacy protection method for sharing social network environment information, which is characterized by comprising:

[0065] Step 1: Establish a cascade scale prediction model for private information at different time points to obtain the propagation scale of private information over time;

[0066] Specifically include:

[0067] Step 11: Structural learning: Based on the graph convolutional network, the contextualized structural patterns in the information cascade graph and the implicit relationships of users in the social network are captured and modeled to obtain the node feature matrix H GCN ; The layer-by-layer propagation rule of GCNs is defined as:

[0068]

[0069] In the formula, H GCN represents the node feature matrix obtained by GCN, represents the adjacency matrix after adding self-loops to ensure that each node includes its own characteristics during the aggregation process, thereby capturing the influence of its direct neighbors and the intrinsic characteristics of the node; yes The degree matrix is ​​used for normalization to avoid thThe layer represents the node features, capturing the properties of the nodes and their neighborhood structures in the layer; H (l) represents the node feature representation of the lth layer, W (l) is a layer-specific trainable weight matrix used to transform the aggregated features, and σ is an activation function that introduces nonlinearity;

[0070] Step 12: Time propagation modeling: Use the self-attention mechanism in Transformer to model the temporal dependency in information propagation. The attention mechanism captures long-term dependencies, improves parallel computing efficiency, and obtains the output result H. Trans ;

[0071]

[0072] Where Q = W Q H GCN ,K=W K H GCN ,V=W V H GCN are the query, key, and value matrices derived from the input sequence, respectively, and W Q ,W K ,W V are trainable weight matrices for query, key, and value, respectively, which compute the relevance (attention score) between different positions in the sequence; d k is the dimension of the key, used to scale the dot product to maintain the stability of the gradient;

[0073] Step 13: Modeling the uncertainty of propagation: Use Bayesian neural networks to model the changes and uncertainties in information propagation and information growth to obtain more robust prediction results. Bayesian neural networks provide a principled approach to dealing with uncertainty by treating network weights as random variables, thereby capturing the variability of user behavior and cascade growth; the prediction distribution of the Bayesian neural network is:

[0074]

[0075] In the formula, H BNN represents the output of the Bayesian neural network, H Trans Represents the output of Transformer, The given data The posterior distribution of the weights represents the weighted belief after observing the data; p(y|H Trans ,W) is the given input H Trans The likelihood of the output with weight W indicates the possibility of the output under a specific weight;

[0076] Step 14, predictor: Combine graph convolutional networks and LSTM to learn high-order representations of information cascades. First, the graph convolutional network is used to capture structural information, then the LSTM is used to capture time series dependencies, and finally the multi-layer perceptron is used to make predictions to obtain the propagation scale of private information over time.

[0077]

[0078] In the formula, Indicates private information I S The propagation scale prediction result at time t, H GCN 、h LSTM and H BNN They represent the outputs of GCN, LSTM, and BNN respectively, concat represents the concatenation operation, and MLP is used for the final prediction.

[0079] Step 2: In online social networks, based on the sensitivity of the privacy attributes contained in the private information and the spread scale obtained in step 1, evaluate the influence of the privacy subject’s private information recipient on the privacy subject.

[0080]

[0081] In the formula, δ is the natural decay rate of influence, which means that influence will naturally weaken over time. That is, when a user speaks less or speaks less frequently in OSNs, his influence will gradually weaken over time. For privacy information S The scale of dissemination in online social networks, that is, the visibility of information. It is private information S The privacy attributes contained in From another perspective, ω represents the user's interest in the private information or the quality of the private information in the online social network. α and β are weight parameters that regulate the impact of information visibility and information quality on influence in the online social network.

[0082] Step 3: Evaluate the privacy risk metric from the perspectives of objective risk and subjective expected risk; where privacy risk R(I S ,t) is:

[0083] R(I S ,t)=r(I S ,t)+N(I S ,t) (6)

[0084] In the formula, r(I S ,t) represents different time t, I S The objective risks faced; N(IS ,t) represents different time t, I S Risk of subjective expectations;

[0085] The objective risks can be assessed from two aspects: the sensitivity and visibility of privacy information:

[0086]

[0087] In the formula, It is private information S The privacy attributes contained in sensitivity, It is private information S the scale of future spread in the current environment;

[0088] When considering privacy risk assessment, the present invention needs to consider the uncertainty factors in the assessment process, namely, subjective factors such as subjective willingness to conceal. Therefore, fuzzy mathematics can be used to describe the uncertainty of privacy risk assessment. The present invention takes subjective willingness to conceal and calculates subjective privacy risk based on fuzzy theory. The corresponding fuzzy set and membership function can be set. Finally, the output logical language value is defuzzified. Since the regional centroid method is applicable to the Mamdani type fuzzy model, the present invention selects the regional centroid method to convert the subjective concealment intention into an accurate subjective privacy risk assessment. Specifically, if the value range of the fuzzy set is [0,1], the calculation formula is as follows:

[0089]

[0090] In the formula, μ is the membership function, Indicates the intention to conceal, r(I S ,t) represents objective risk.

[0091] Step 4: Determine the privacy gain based on the influence obtained in step 2, determine the privacy loss based on the privacy risk obtained in step 3, and determine the privacy utility based on the privacy gain and privacy loss;

[0092] Privacy benefits of information sharing: To describe the impact of social benefits on individual decisions, we use the term influence to model the social benefits that individuals gain when they decide to share information with an audience.

[0093] Privacy loss of sharing information: Privacy loss refers to an individual’s assessment of privacy leakage in the process of deciding whether to share personal information. Therefore, the privacy risk value can be directly used to assess privacy loss; the privacy cost of individual S is

[0094] Privacy benefits of not sharing information: Since online social networks are mainly social platforms for communication between users, it is assumed that when an individual chooses not to share personal information, the social benefits at the current moment, i.e., influence

[0095] Privacy loss of not sharing information: Similar to the privacy loss of sharing information, the privacy loss of not sharing information is determined using the foreground privacy risk. However, in this case, the privacy risk value is zero and the privacy loss is

[0096] From the above analysis, we can see that the utility of sharing information is the sum of social benefits and privacy losses, that is, in is a negative value, indicating a personal loss. At the same time, the utility of not sharing information is only determined by the current social benefit and the privacy loss of the previous moment The sum of Therefore, the decision of individual S on whether to share information with the audience can be determined by measuring the utility value, and the utility function is as follows:

[0097]

[0098] In the formula, is the privacy loss. At different times, when the privacy information of the privacy subject S is propagated in OSNs, σ S =1, otherwise σ S =0.

[0099] Step 5: The privacy utility maximization problem in online social networks is summarized as a constrained Markov decision process to analyze the decision benefits at different time points and develop appropriate strategies to achieve balance;

[0100] Specifically include:

[0101] Step 51: In the Markov decision process, the reward function is determined by the current state of the system and the selected action. The utility function U(S) is selected as the reward function in the Markov decision process. The utility function considers the social benefits and privacy losses when sharing and not sharing information, and changes with the choice of the privacy subject S at different decision moments. S and change;

[0102] Step 52: Relative value iteration (RVI) is used to solve the optimal strategy. First, the Lagrangian method is used to transform the CMDP problem into an unconstrained MDP problem. The utility function U(S) and the risk function R(I) are transformed by introducing the Lagrangian multiplier β (β>0). S,t) to form a Lagrangian reward function L(S,t) to solve the optimal strategy; the above Lagrangian reward function takes into account the balance between privacy utility and risk, and then solves the optimal strategy through the relative value iteration method;

[0103] Step 53: When the CMDP problem is converted into an unconstrained MDP problem, the reward function L(S, t) is used as the objective function for optimization; at this time, the constraint condition (i.e., the maximum risk R that the privacy subject can bear) max ) by R(I S ,t) Determine that there is a relationship between the Lagrange multiplier β and the maximum risk; Use the Q-learning algorithm to determine the maximum risk R that the privacy subject can bear max and the corresponding Lagrange multiplier β * , and through β * Control the balance between privacy risk and utility, thereby optimizing the reward function under unconstrained MDP problems and ultimately solving the optimal strategy;

[0104] Step 54: To obtain the optimal strategy, firstly perform Lagrange multiplier β * Perturb and get β - =β * -Δβ and β + =β * +Δβ; then the perturbed β - and β + Substitute the reward function L(S,t) and solve the corresponding two pure strategies δ by relative value iteration method - and δ + , and combined with the utility function U(S) and the risk function R(I S ,t) respectively calculate the average risk R - and R + ; Since the optimal strategy is a mixed strategy of two pure strategies, the present invention defines the parameter q so that Based on the specific value of q obtained from this relationship, the optimal strategy can be obtained through a comprehensive balance of utility and risk at each decision moment:

[0105]

[0106] like Figure 1 As shown, the present invention provides a privacy protection system for sharing social network environment information, including:

[0107] A private information propagation scale prediction module, used to implement step 1 of the above privacy protection method;

[0108] The influence measurement and update module is used to implement step 2 of the above privacy protection method;

[0109] A privacy risk measurement module, used to implement step 3 of the above privacy protection method;

[0110] A privacy utility measurement module, used to implement step 4 of the above privacy protection method;

[0111] The privacy policy decision module is used to implement step 5 of the above privacy protection method.

[0112] The advantages of the present invention are:

[0113] 1. At present, the information sharing privacy protection technology in online social networks mainly formulates privacy protection strategies through static information and user personal information, ignoring the dynamically changing user roles, user relationships and information propagation in online social networks; the present invention effectively formulates a privacy protection strategy that conforms to the dynamic propagation of privacy information by introducing the privacy information propagation characteristics in online social networks.

[0114] 2. The present invention evaluates the influence of privacy subjects, update methods and privacy sharing risks through the scale of privacy information dissemination, and effectively evaluates the privacy utility that meets the personalized needs of users.

[0115] 3. The present invention maximizes the privacy utility obtained by sharing the benefits and losses of private information by constraining the Markov decision process, thereby formulating a privacy protection strategy that meets the personalized needs of users and achieving the goal of balancing privacy protection and privacy sharing.

[0116] The above are only preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A privacy protection method for information sharing in a social network environment, characterized in that: include: Step 1: Establish a cascade scale prediction model for private information at different time points to obtain the propagation scale of private information over time; Step 2: Based on the sensitivity of the privacy attributes contained in the private information and the propagation scale obtained in step 1, evaluate the influence of the privacy subject's private information receiver on the privacy subject; Step 3: Evaluate privacy risks from the perspectives of objective risks and subjective expected risks; Step 4: Determine the privacy gain based on the influence obtained in step 2, determine the privacy loss based on the privacy risk obtained in step 3, and determine the privacy utility based on the privacy gain and privacy loss; Step 5: The privacy utility maximization problem in online social networks is summarized into a constrained Markov decision process to analyze the decision benefits at different time points and formulate appropriate strategies to achieve balance. In the Markov decision process, the reward function of one step is determined by the current state of the system and the selected action. The privacy utility obtained in step 4 is selected as the reward function in the Markov decision process.

2. The privacy protection method for sharing social network environment information as claimed in claim 1, characterized in that: The step 1 specifically includes: Step 11: Based on the graph convolutional network, the contextualized structural patterns in the information cascade graph and the implicit relationships of users in the social network are captured and modeled to obtain the node feature matrix H GCN ; Step 12: Use the self-attention mechanism in Transformer to model the temporal dependency in information propagation and obtain the output result H Trans ; Step 13: Use Bayesian neural network to model the changes and uncertainties in information propagation and information growth, and obtain the output result H BNN ; Step 14: First, capture the structural information through the graph convolutional network, then capture the time series dependency through the LSTM, and finally make predictions through the multi-layer perceptron to obtain the propagation scale of the private information over time.

3. The privacy protection method for sharing social network environment information as claimed in claim 2, characterized in that: In step 11, H GCN The calculation formula is: In the formula, H GCN represents the node feature matrix obtained by GCN, represents the adjacency matrix after adding the self-loop, yes The degree matrix of (l) represents the node feature representation of the lth layer, W (l) is a layer-specific trainable weight matrix, and σ is an activation function that introduces nonlinearity; In step 12, H Trans The calculation formula is: Where Q = W Q H GCN ,K=W K H GCN ,V=W V H GCN are the query, key, and value matrices derived from the input sequence, respectively, and W Q ,W K ,W V are trainable weight matrices for query, key, and value, respectively; d k is the dimension of the key; In step 13, H BNN The calculation formula is: In the formula, H BNN represents the output of the Bayesian neural network, H Trans Represents the output of Transformer, The given data The posterior distribution of the weights, p(y|H Trans ,W) is the given input H Trans and the likelihood of the output of weight W; In step 14, the private information I S Prediction results of the spread scale at time t The calculation formula is: In the formula, Indicates private information S The propagation scale prediction result at time t, H GCN 、h LSTM and H BNN They represent the outputs of GCN, LSTM, and BNN respectively, concat represents the concatenation operation, and MLP is used for the final prediction.

4. The privacy protection method for sharing social network environment information as claimed in claim 3, characterized in that: In step 2, the influence The calculation formula is: Where δ is the natural attenuation rate of influence, For privacy information S The scale of spread in online social networks, It is private information S The privacy attributes contained in is the sensitivity of , α and β are weight parameters.

5. The privacy protection method for sharing social network environment information as claimed in claim 4, characterized in that: In step 3, the privacy risk R(I S ,t) is: R(I S ,t)=r(I S ,t)+N(I S ,t) (6) In the formula, r(I S ,t) represents different time t, I S The objective risks faced; N(I S ,t) represents different time t, I S Risk of subjective expectations; In the formula, It is private information S The privacy attributes contained in sensitivity, It is private information S the scale of future spread in the current environment; In the formula, μ is the membership function, Indicates the intention to conceal, r(I S ,t) represents objective risk.

6. The privacy protection method for sharing social network environment information as claimed in claim 5, characterized in that: In step 4, the calculation formula of the privacy utility U(S) is: In the formula, is the privacy loss. At different times, when the privacy information of the privacy subject S is propagated in OSNs, σ S =1, otherwise σ S =0.

7. The privacy protection method for sharing social network environment information as claimed in claim 6, characterized in that: The step 5 specifically includes: Step 51: In the Markov decision process, the reward function is determined by the current state of the system and the selected action. The utility function U(S) is selected as the reward function in the Markov decision process. The utility function considers the social benefits and privacy losses when sharing and not sharing information, and changes with the choice of the privacy subject S at different decision moments. S and change; Step 52: Relative value iteration is used to solve the optimal strategy. First, the Lagrangian method is used to transform the CMDP problem into an unconstrained MDP problem. The utility function U(S) and the risk function R(I) are transformed by introducing the Lagrangian multiplier β. S ,t) to form a Lagrangian reward function L(S,t) to solve the optimal strategy; the above Lagrangian reward function takes into account the balance between privacy utility and risk, and then solves the optimal strategy through the relative value iteration method; Step 53: When the CMDP problem is converted into an unconstrained MDP problem, the reward function L(S, t) is used as the objective function for optimization; at this time, the constraints are R(I S ,t) Determine that there is a relationship between the Lagrange multiplier β and the maximum risk; Use the Q-learning algorithm to determine the maximum risk R that the privacy subject can bear max and the corresponding Lagrange multiplier β * , and through β * Control the balance between privacy risk and utility, thereby optimizing the reward function under unconstrained MDP problems and ultimately solving the optimal strategy; Step 54: To obtain the optimal strategy, firstly perform Lagrange multiplier β * Perturb and get β - =β * -Δβ and β + =β * +Δβ; then the perturbed β - and β + Substitute the reward function L(S,t) and solve the corresponding two pure strategies δ by relative value iteration method - and δ + , and combined with the utility function U(S) and the risk function R(I S ,t) respectively calculate the average risk R - and R + ; Since the optimal strategy is a mixed strategy of two pure strategies, the optimal strategy can be obtained through a comprehensive balance of utility and risk at each decision moment: In the formula, 8. A privacy protection system for sharing social network environment information, used to implement the privacy protection method for sharing social network environment information as described in any one of claims 1 to 7, characterized in that: include: A private information propagation scale prediction module is used to implement step 1 of the privacy protection method; The influence measurement and update module is used to implement step 2 of the privacy-preserving method; A privacy risk measurement module, used to implement step 3 of the privacy protection method; The privacy utility measurement module is used to implement step 4 of the privacy protection method; The privacy policy decision module is used to implement step 5 of the privacy protection method.

Images