Personalized sampling graph collection with local differential privacy for link prediction

Link prediction (LP) is an attractive research problem on social network data. Yet, the link information may be leaked by untrusted collector. As countermeasures, there are a few methods specially designed for under local differential privacy (LDP), which allow third-party data collector collects user while protecting connection privacy. In this paper, we propose C ommunity-based graph collection with P ersonalized sampling R andomized esponse (CPRR) novel algorithm LDP and reaches decent trade-off between sensitive protection performance. The proposed mechanism adopts personalized technique each of then utlizes randomized response sampled subset. Based technique, can reduce injected noise in LDP. Meanwhile, considering different edge distributions regions original graph, community-based strategy. Then, prove that CPRR satisfies Through extensive experiments several real-life datasets, demonstrate achieve better results balancing performance than state-of-art baselines.