ALPINE: A Lightweight and Adaptive Privacy-Decision Agent Framework for Dynamic Edge Crowdsensing

TL;DR

ALPINE addresses dynamic privacy risks in MECS by combining on‑device multi‑dimensional risk perception with a TD3‑based privacy budget controller and edge‑side verification. The framework integrates LightAE for channel risk, semantic and contextual risk modeling, and a bounded Laplace mechanism for DP noise, all within a closed‑loop system that continually refines policy based on edge feedback. Theoretical results on DP composition and optimality, together with extensive experiments on real hardware and datasets, demonstrate improved privacy resilience while maintaining data utility and low on‑device overhead, supporting scalable deployment in large heterogeneous edge environments. This work provides a practical, adaptive privacy solution for real‑time edge sensing, balancing protection, data usefulness, and resource constraints in dynamic urban/industrial settings.

Abstract

Mobile edge crowdsensing (MECS) systems continuously generate and transmit user data in dynamic, resource-constrained environments, exposing users to significant privacy threats. In practice, many privacy-preserving mechanisms build on differential privacy (DP). However, static DP mechanisms often fail to adapt to evolving risks, for example, shifts in adversarial capabilities, resource constraints and task requirements, resulting in either excessive noise or inadequate protection. To address this challenge, we propose ALPINE, a lightweight, adaptive framework that empowers terminal devices to autonomously adjust differential privacy levels in real time. ALPINE operates as a closed-loop control system consisting of four modules: dynamic risk perception, privacy decision via twin delayed deep deterministic policy gradient (TD3), local privacy execution and performance verification from edge nodes. Based on environmental risk assessments, we design a reward function that balances privacy gains, data utility and energy cost, guiding the TD3 agent to adaptively tune noise magnitude across diverse risk scenarios and achieve a dynamic equilibrium among privacy, utility and cost. Both the collaborative risk model and pretrained TD3-based agent are designed for low-overhead deployment. Extensive theoretical analysis and real-world simulations demonstrate that ALPINE effectively mitigates inference attacks while preserving utility and cost, making it practical for large-scale edge applications.

Figures (8)

• Figure 1: We design an adaptive lightweight privacy-protection agent framework ALPINE. A closed-loop control process: (1) Server launches task; (2) Environmental risk score is forwarded to decision agent; (3) Noise is injected according to the decision; (4) Processed data are transmitted to the server for validation; (5) Validation results are fed back.
• Figure 2: LightAE with block-granularity scaling.
• Figure 3: The comparison of three RL models.
• Figure 4: The performance of TD3 model
• Figure 5: Evaluation in Intel Berkeley Research Lab Sensor Data