A joint privacy protection algorithm for edge computing task offloading based on Dempster–Shafer evidence theory

• Published in: Physical communication Vol. 72; p. 102813
• Main Authors: Chen, Tian, Hao, Wanming, Yang, Wencong, Yang, Shouyi, Sun, Xuandi, Tang, Zhiqing
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 01.10.2025
• Subjects: Edge computing; Evidence theory; Internet of things; Privacy; Privacy; Internet of things; Evidence theory; Edge computing
• ISSN: 1874-4907
• DOI: 10.1016/j.phycom.2025.102813

Edge computing reduces system latency by offloading tasks from cloud servers to Internet of Things (IoT) devices. This architecture decreases network congestion, improves system efficiency, and enables real-time processing. However, in consideration of resource efficiency, user devices tend to select the nearest cloud servers, which may enable potential malicious eavesdropping devices to infer user location through intercepted communication information. Existing research has used privacy entropy to measure user location privacy, and assigned different privacy entropy values to various types of cloud servers. However, user privacy protection in real-world scenarios depends on multiple factors, not just the classification of edge cloud servers. To solve this problem, we propose a joint privacy protection algorithm (JPPA), which introduces a novel metric for user location privacy guarantee called belief, this metric combines two key factors: task caching status and the geographical location of edge cloud servers. We formulate the problem as a constrained Markov decision process (CMDP), aiming to minimize the average network traffic subject to constraints on belief levels and energy outage probability. By transforming the CMDP into a linear programming problem, computational feasibility is ensured. Simulation results demonstrate that compared to traditional privacy entropy-constrained methods, JPPA reduces average network traffic overhead more effectively while ensuring user location privacy.