Abstract:With the emergence of network-centric data, social network graph publishing is conducive to data analysts to mine the value of social networks, analyze the social behavior of individuals or groups, implement personalized recommendations, and so on. However, published social network graphs are often subject to re-identification attacks from adversaries, which results in the leakage of users’ privacy. The k -anonymity technology is widely used in the field of graph publishing, which is quite effective to resist re-identification attacks. However, the current researches still exist some issues to be solved: the protection of directed graphs is less concerned than that of undirected graphs; the protection of graph structure is often ignored while achieving the protection of nodes’ identities; the same protection is performed for different users, which doesn’t meet the different privacy requirements of users. Therefore, to address the above issues, a multi-level k -degree anonymity (MLDA) scheme on directed social network graphs is proposed in this paper. First, node sets with different importance are divided by the firefly algorithm and constrained connectedness upper approximation, and they are performed different k -degree anonymity protection to meet the different privacy requirements of users. Second, a new graph anonymity method is proposed, which achieves the addition and removal of edges with the help of fake nodes. In addition, to improve the utility of the anonymized graph, a new edge cost criterion is proposed, which is used to select the most appropriate edge to be removed. Third, to protect the community structure of the original graph as much as possible, fake nodes contained in a same community are merged prior to fake nodes contained in different communities. Experimental results on real datasets show that the newly proposed MLDA scheme is effective to balance the privacy and utility of the anonymized graph.

Protecting sensitive labels in social network data anonymization

Protecting Sensitive Labels in Weighted Social Networks.

Privacy Preservation in Social Network Against Public Neighborhood Attacks

Preserving Privacy In Social Networks Against Label Pair Attacks

Protect You More Than Blank：Anti-Learning Sensitive User Information in the Social Networks

Distributed and Personalised Social Network Privacy Protection

De-Anonymizing Dynamic Social Networks.

Against Signed Graph Deanonymization Attacks on Social Networks

On Anonymization of Multi-graphs.

Structure-Attribute-Based Social Network Deanonymization with Spectral Graph Partitioning

MLDA: a Multi-Level K -Degree Anonymity Scheme on Directed Social Network Graphs

SA Framework Based De-anonymization of Social Networks

Sensitive Labels Matching Privacy Protection in Multi-Social Networks.

On the privacy and utility of anonymized social networks

Sensitive Label Privacy Preservation with Anatomization for Data Publishing

Graph Publication when the Protection Algorithm is Available

Anonymization in Online Social Networks Based on Enhanced Equi-Cardinal Clustering.

Anonymizing Hypergraphs with Community Preservation

PLRD-(k,m):Distributed k-Degree-m-Label Anonymity with Protecting Link Rela-tionships??

Multi-View Low-Rank Coding-Based Network Data De-Anonymization

Splitting Anonymization: a Novel Privacy-Preserving Approach of Social Network