计算机工程与应用2026，Vol.62Issue(17)：306-317,12.DOI:10.3778/j.issn.1002-8331.2508-0215
本地差分隐私下能力感知的动态数据收集机制研究
Capacity-Aware Dynamic Data Collection Mechanism Under Local Differential Privacy
摘要
Abstract
Mobile crowdsensing is widely applied in various data collection scenarios.Among these,dynamic scenarios where tasks arrive online one by one and users'locations are constantly changing pose significant challenges,and this process also involves privacy risks.Most existing works only focus on a single task allocation algorithm and adopt a uni-fied privacy strategy.However,they fail to properly balance efficiency and privacy when considering the heterogeneity in users'capabilities and privacy demands,especially when users'capabilities are dynamically changing.To address this,the paper proposes a capability-aware data collection mechanism based on a two-stage algorithm under local differential privacy.Firstly,the QUCB-PC method is proposed,which integrates users'capabilities and privacy demands into a repre-sentation of data quality.Combined with the multi-armed bandit model,it evaluates users'potential value for task allo-cation and alleviates the interference of Laplacian noise in location information through probability comparison.Then,the EM framework is adjusted by introducing prior information such as users'capabilities,and the MAP-EM method is proposed to infer the current users'capabilities and task ground truths in parallel,while suppressing estimation bias caused by Gaussian noise in sensed data.In terms of privacy,the noise magnitude is flexibly controlled according to privacy requirements.Simulation results show that the proposed methods perform well in terms of cumulative joint benefit,average task quality,and root mean square error,with significant improvements compared with benchmark methods.关键词
移动群智感知(MCS)/本地差分隐私(LDP)/用户能力/多臂赌博机/EM算法Key words
mobile crowdsensing(MCS)/local differential privacy(LDP)/user capability/multi-armed bandit/expectation-maximization(EM)algorithm分类
信息技术与安全科学引用本文复制引用
陈家庆,方贤进,王杰,张朋飞..本地差分隐私下能力感知的动态数据收集机制研究[J].计算机工程与应用,2026,62(17):306-317,12.基金项目
国家自然科学基金(61572034). （61572034）
