通信学报2026，Vol.47Issue(8)：60-77,18.DOI:10.11959/j.issn.1000-436x.TXXB260318
数据流场景下内存受限本地差分隐私协议鲁棒性评估与增强
Evaluating and enhancing the robustness of memory-bounded LDP protocols for streaming data
摘要
Abstract
Local differential privacy(LDP)is vulnerable to data poisoning attacks,yet the robustness of memory-bounded streaming data protocols with domain compression remains understudies.Four mainstream LDP streaming fre-quent item identification protocols(BGR,DSR,BDR,and CNR)were investigated.Firstly,an attack-driven evaluation framework was proposed,and a data poisoning attack maximizing frequency gain was designed.Secondly,using attack success rate(ASR)as the evaluation metric,experiments on three datasets showed that DSR was the most robust proto-col,and storage size significantly impacts robustness.Finally,a protocol-agnostic attack detection scheme based on dis-tribution consistency was proposed,utilizing data frequency reconstruction to address missing global frequencies.Experi-mental results demonstrate that the proposed scheme achieves significant detection performance under low privacy bud-gets or high poisoning ratios.关键词
本地差分隐私/鲁棒性评估/数据流挖掘/内存受限/数据投毒Key words
local differential privacy/robustness evaluation/data stream mining/memory bounded/data poisoning分类
信息技术与安全科学引用本文复制引用
田月池,彭锦钰,李效光,张芊龙,李凤华,左金鑫,牛犇..数据流场景下内存受限本地差分隐私协议鲁棒性评估与增强[J].通信学报,2026,47(8):60-77,18.基金项目
国家自然科学基金资助项目(No.62332018,No.62402363) （No.62332018,No.62402363）
陕西省博士后科研项目资助(No.2024BSHSDZZ106) The National Natural Science Foundation of China(No.62332018,No.62402363),Shaanxi Province Postdoc-toral Research Funding Program(No.2024BSHSDZZ106) （No.2024BSHSDZZ106）
