POTENTIAL: Privacy-Oriented Task Offloading Via Context-Aware Risk Estimation in JCAS-Enabled Edge Networks

Joint Communication and Sensing (JCAS) enables User Equipment (UE) to offload computational tasks to nearby peer devices or edge servers over context-aware wireless links. While this improves latency and resource utilization, the dual usage of the wireless spectrum for communication and environmental sensing introduces novel privacy risks, particularly for tasks involving location-sensitive or contextual data. This paper introduces a novel offloading framework that leverages real-time environmental information sensed by JCAS base stations to dynamically estimate the privacy risk associated with each offloading link. Based on these context-aware risk estimates, UE selects among local, Device-to-Device (D2D), or edge execution modes to minimize a composite cost function balancing privacy leakage and delay. A lightweight greedy scheduler is applied to solve the offloading decision problem efficiently under deadline constraints. Simulation results demonstrate that our method significantly reduces effective privacy leakage with negligible latency overhead, consistently outperforming context-agnostic baseline strategies.