【摘要】
群智感知服务的广泛应用带来了个人隐私的泄漏,然而现存的隐私保护策略不能适应群智感知环境.针对相关缺陷,提出了一种移动群智感知中基于深度强化学习的隐私保护策略.该策略通过泛化任务,使得攻击者无法分辨用户具体完成了哪一个任务,切断了用户和任务之间的关联,保护了用户的位置隐私.当混淆任务数量不足以达到用户的隐私保护需求时,使用抑制法放弃该任务.该策略使用深度强化学习的方法不断尝试不同的混淆任务组合,训练一个可以输出最低抑制率的混淆任务选择方案的深度Q网络.实验结果表明,上述策略在不破坏感知任务有效性的前提下,以较低的抑制率保护了用户的位置隐私.%The widespread use of Crowdsensing has brought about personal privacy disclosure. However,the existing privacy protection strategy are not directly tailored for Crowdsensing applications. To address this issue,this paper proposes a deep reinforcement learning based privacy protection strategy. The strategy protects privacy by generalizing Crowdsensing tasks,which cuts off the association between users and tasks,makes it impossible for attackers to distinguish which task was completed by the target user. The task is suppressed when the number of obfuscation tasks is not enough to meet the user's privacy protection requirements. This strategy uses a deep reinforcement learning algorithm to tries different anonymity task set,training a deep Q-network that can output a anonymity task set with the minimum suppression rate. The experimental results demonstrate that the proposed strategy protect users' privacy effectively while ensuring data validity and achieve significant performance improvement in terms of suppression rate.