Research On Community Privacy Protection Algorithm Based On Graph Structure Perturbation

Community detection technology,as an important branch of community research,plays an important role in social networks,bioinformatics,business data analysis.The in-depth analysis of network community structures by the technology also puts potential user information in plain sight.Attackers can infer private information such as user preferences through community detection,leading to privacy leakage,and community privacy protection technology has emerged.The current community privacy protection algorithms suffer from low data utility due to the destruction of important network characteristics such as node reachability when perturbing the network structure,waste of computational resources due to over-protection of common nodes,high cost of practical implementation of some perturbation operations,and inability to resist compound background knowledge attacks.In this thesis,two community privacy protection algorithms are proposed to address the different requirements for community privacy protection in different application scenarios:A hierarchical community privacy protection algorithm based on selective edge replacement is proposed for the problem that some community privacy protection algorithms destroy node reachability and overprotect common nodes.The algorithm achieves more efficient graph structure perturbation and reasonable degree of privacy protection by dividing node level and edge level and calculating edge perturbation and retention probability.Meanwhile,in order to ensure node accessibility,the algorithm selects the L-hop candidate neighbor node of the deleted edge node as the relay node,and selectively replaces the deleted edge.The experimental results show that the algorithm improves the availability of data while ensuring that the privacy protection strength of the community is not reduced.A K-Degree-L-Neighbor anonymity-based community privacy protection algorithm is proposed for the problem that the edge deletion operation has low feasibility in application scenarios such as attention networks and cannot resist compound background knowledge attacks.The algorithm only adds points and edges to the network structure,drawing lessons from the idea of anonymity to resist node-degree background knowledge attacks,while limiting the data loss due to large changes in the graph structure.The algorithm incorporates a certain degree of randomness in node selection to make it more difficult for attackers to crack and achieve effective hiding of community structure as well as node degree.The experimental results show that this algorithm outperforms the comparison algorithms in community hiding and can resist the degree background knowledge attack.