一种MEC中隐私保护计算卸载方法

Privacy-Preserving Computation Offloading Method of Multi-Access Edge Computing

针对多接入边缘计算(Multi-access Edge Computing,MEC)中用户计算卸载决策可能暴露用户特征泄露用户隐私的问题,提出一种基于卸载频率的隐私保护计算卸载方法。首先,分析了计算任务卸载频率的隐私暴露风险,并在现有卸载决策模型中引入隐私限制使各计算任务卸载频率尽可能偏离其原始卸载频率;然后,提出了假任务机制权衡终端能耗和隐私保护的关系,当因隐私限制无法正常执行计算卸载时,在MEC节点生成虚假的卸载任务增大隐私保护效果;最后,基于贪心思想简化模型,提出基于神经网络的隐私保护卸载算法(ANN based Privacy Preserving Offloading Algorithm,APPOA)求解。仿真结果表明,基于卸载频率的隐私保护计算卸载方法能改变特征任务的卸载频率以满足隐私限制,且平均能耗降低了至少65%。

The computation offloading decision in multi-access edge computing(MEC)may expose user’s characteristics and cause the user to be locked.In this paper a privacy-preserving computation offloading method based on offloading frequency is proposed.Firstly,privacy threat of tasks’offloading frequencies is analyzed,and privacy restriction is added to current offloading decision model to amplify the deviation of each task’s offloading frequency;then,the chaff task mechanism is proposed to balance the terminal energy consumption and privacy preservation,making the task with the highest offloading frequency to meet the privacy restriction by generating a same fake task on MEC node when offloading is not performed due to privacy restrictions;finally,the privacy-preserving computation offloading decision is modeled and solved based on artificial neural networks.Simulation results validate that the privacy-preserving computation offloading method based on offloading frequency can change the tasks’offloading frequencies to meet privacy restrictions,and the average energy consumption is reduced by at least 65%.