The Least Information Set Needed by Privacy Attackers

• 成果类型: Article
• 署名作者: Xie, Antai; Wang, Xiaofan; Yang, Wen; Cao, Ming; Ren, Xiaoqiang
• 署名单位: Shanghai University; East China University of Science & Technology; University of Groningen; Shanghai University
• 刊物名称: IEEE TRANSACTIONS ON AUTOMATIC CONTROL
• ISSN/ISSBN: 0018-9286
• DOI: 10.1109/TAC.2024.3470848
• 发表日期: 2025
• 页码: 1652-1666
• 关键词: Consensus algorithm; Directed graphs; noise; accuracy; vectors; CONVERGENCE; Matrix decomposition; Differential privacy; Heuristic algorithms; Eavesdropping; Average consensus; information set; privacy-preserving

摘要

This article studies privacy-preserving problems focusing on the information that is available to the attacker. For a class of general linear average consensus algorithms, we identify the least information set required by an attacker to infer the initial value of the target node. We then propose a novel algorithm that protects the initial value effectively if the attacker fails to have this least information. Furthermore, we show that the initial value of nodes may be disclosed even if the attacker does not know any information exchanged between the target node and its neighbors. Finally, we propose a novel eavesdropping algorithm if the least information is available. Several numerical examples are given to verify the validity of our results.