Research On Privacy Attribute Inferring Attack Based On Data Mining In Social Network

With the rapid development of social networks,there are many techniques for reasoning attacks against the privacy attributes of social network users.The attacker uses the publicly visible attributes and social network structure of the user to infer the missing or incomplete attribute data of the user.The current research methods mainly include attribute-based reasoning attacks based on user behavior and attribute-based reasoning attacks based on users'social friends.The current two methods do not consider the inherent association weight between user attributes and attributes.This paper makes the following research on the above issues:First of all,this paper proposes an attribute reasoning attack framework based on attribute weights and social structure.This paper firstly combines the social network and attribute information to mine the relationship and weight between the attributes,and proposes a user attribute reasoning attack framework combining the attribute weight,social structure and user public attributes.Secondly,this paper implements the algorithm of the above attack framework.The FP-Growth algorithm is implemented in the attribute weight mining stage to derive the weight value between attribute pairs.In the attribute reasoning attack stage,the probability-based attack algorithm is implemented.The candidate attribute set of the userperforms probability calculation for each candidate attribute,and finally selects k candidate attributes with the highest probability value as the inference result.In each candidate attribute probability calculation process,the user who shares the attribute with the target user is first divided according to the public attribute of the target user and the candidate attribute(6₄),and is divided into a user set that shares different attribute values with the target user,and each set is shared.The candidate attribute occurrence probability is calculated by weighted sum,and the weighted value is divided into three,the attribute contact weight₁,the target user social weight₂,and the target user share attribute weight₃.The algorithm in this paper is also applicable to the scenario of speculating whether the information disclosed by the user is false information.Similarly,this paper can calculate the probability of occurrence of a specific attribute.When the probability is lower than the threshold,the specific threshold is determined according to the experimental results.The disclosed attributes may be false information provided by the user to protect personal privacy.Finally,the effectiveness of the attack algorithm is verified by design experiments on Google+dataset.The experimental results show that the WASN attribute inference algorithm is effective in different kinds of attribute speculation.The experimental comparison results in this paper show that it is better than other algorithms.