核心概念
Proposing a novel k-stars LDP algorithm for (p, q)-clique enumeration with improved utility and privacy protection.
要約
The article introduces the concept of k-stars LDP as a novel framework for (p, q)-clique enumeration under local differential privacy. It addresses the limitations of traditional edge LDP algorithms by reducing estimation errors and improving data utility. The proposed algorithm utilizes k-stars neighboring lists, absolute value correction technique, and k-stars sampling technique to enhance privacy protection while counting subgraphs. The theoretical analysis and experiments demonstrate the superiority of the k-stars LDP algorithm over edge LDP algorithms in terms of privacy, unbiasedness, and utility. The experiments conducted on real datasets show better performance of k-stars LDP in sparse graphs and for different (p, q)-cliques.
統計
"The number of (2, 2)-clique and (1, 2)-clique in Gplus and Facebook grows as n increases."
"The computational complexity of L2 loss can be expressed as O(f22(G)+f12(G))."
"The number of (2, 2)-clique is larger than that of (1, 2)-clique in dense graphs like Gplus and Facebook."
"The L2 loss of k-stars LDP in GitHub is smaller than that of the rest datasets."
"The computational complexity of relative error is O(qf12(G)/f22(G))."
引用
"Our proposed k-stars LDP algorithm has a better utility than traditional edge LDP algorithm."
"The proposed k-stars LDP algorithm can handle complex subgraph enumeration problems better than traditional edge LDP algorithm."
"The k-stars LDP algorithm works better in a sparse graph, which means it can work better in the practical scenario as real social network tends to be sparse."