当前位置: X-MOL 学术IEEE Wirel. Commun. › 论文详情
Our official English website, www.x-mol.net, welcomes your feedback! (Note: you will need to create a separate account there.)
Learning-Empowered Privacy Preservation in Beyond 5G Edge Intelligence Networks
IEEE Wireless Communications ( IF 10.9 ) Pub Date : 2021-05-14 , DOI: 10.1109/mwc.001.2000331
Pengcheng Zhu , Jun Xu , Jiamin Li , Dongming Wang , Xiaohu You

Edge intelligence is a key enabler for powerful applications through offering artificial-intelligence-empowered computation and caching service in proximity of end users. However, massive data interaction among edge nodes and distributed information storage in the service process make edge network vulnerable to privacy threats. Moreover, the upcoming beyond fifth generation (B5G) era brings new features for edge services in terms of system complexity and operation mode, which pose significant challenges to user privacy protection. To address the challenges, we first investigate privacy concerns in B5G edge intelligence networks and design privacy-driven application clarification strategies. In order to meet the different privacy requirements of various edge applications, we propose a learning-empowered privacy-preserving scheme, which adaptively applies data perturbation in a multi-mode differential privacy (DP) approach. Then we present a case study that implements our proposed schemes in edge caching service management. Numerical results demonstrate that the schemes efficiently improve edge service utility without loss of privacy.

中文翻译:


超越 5G 边缘智能网络中的学习赋能隐私保护



边缘智能通过在最终用户附近提供人工智能支持的计算和缓存服务,成为强大应用程序的关键推动者。然而,边缘节点之间的海量数据交互以及服务过程中的分布式信息存储使得边缘网络容易受到隐私威胁。此外,即将到来的超第五代(B5G)时代给边缘服务带来了系统复杂性和运行模式方面的新特点,这对用户隐私保护提出了重大挑战。为了应对这些挑战,我们首先调查 B5G 边缘智能网络中的隐私问题,并设计隐私驱动的应用澄清策略。为了满足各种边缘应用的不同隐私要求,我们提出了一种学习授权的隐私保护方案,该方案在多模式差分隐私(DP)方法中自适应地应用数据扰动。然后我们提出一个案例研究,在边缘缓存服务管理中实施我们提出的方案。数值结果表明,该方案有效地提高了边缘服务效用而不损失隐私。
更新日期:2021-05-14
down
wechat
bug