Proceedings of the ACM Turing Celebration Conference - China | 2019
De-anonymize social network under partial overlap
Abstract
De-anonymizing social networks has become a direct threat to people s privacy . Actually, It can be boiled down to graph matching problems. The attackers steal users real information in anonymized social networks by mapping them to secondary cross-domain networks. In particular, when partial node identity in the anonymized network is known, such attacks will become more powerful. Some scholars have studied seeded network de-anonymization. However, there is a lack of consideration for network overlap. We further expand the work of predecessors and consider partially overlapping networks de-anonymization with the aid of seeded nodes. We give a more general form of theoretical results under Erdös - Rényi model(ER model). We also validated our results on both synthetic and real data.