Privacy-aware task scheduling in satellite edge computing: a game-theoretic framework

The growing deployment of low Earth orbit (LEO) constellations equipped with onboard edge computing is making satellite edge computing (SEC) increasingly practical by enabling onboard processing for latency-sensitive tasks. In this setting, heterogeneous onboard computing capacities, bandwidth-limited connectivity, and resource contention make task scheduling indispensable for coordinating task placement and multi-hop forwarding. Conventional schedulers often optimize latency and energy first (tight deadlines, limited onboard energy, and measurable costs), whereas privacy is harder to quantify and thus under-modeled, potentially leading to higher exposure via fast multi-hop forwarding, untrusted execution choices, task co-location, etc. To address these conflicting objectives, this paper formulates a unified optimization framework and proposes two complementary scheduling schemes. We first introduce Task Scheduling Optimizer (TS-OPT), a centralized mixed-integer programming approach that provides an optimal reference for the proposed model. To overcome the scalability limitations of centralized solvers, we subsequently develop the Privacy-Aware Potential Game Scheduler (PAPGS), a decentralized algorithm that casts the problem as an exact potential game and converges in a finite number of best-response updates to a pure-strategy Nash equilibrium. Simulation results demonstrate that PAPGS achieves a 0.35% optimality gap to TS-OPT in small-scale scenarios, and consistently attains lower unified system cost with high deadline satisfaction at larger scales.