Quantifying the Privacy-Utility Trade-off in GPS-based Daily Stress Recognition using Semantic Features

2025-11-28

AI Summary

Balancing privacy preservation and model utility remains challenging in GPS trajectory-driven stress recognition. Method: This paper proposes an end-to-end semantic location encoding framework: it constructs a static semantic map using a self-hosted OpenStreetMap (OSM) engine and performs on-device trajectory anonymization via large language model–guided stop detection and semantic labeling. Contribution/Results: We first quantify the impact of semantic features on the privacy–utility trade-off, demonstrating that under strong privacy guarantees—i.e., no exposure of raw coordinates—the model’s stress recognition performance is statistically indistinguishable from a non-private baseline (p > 0.05). Experiments employ leave-one-subject-out (LOSO) cross-validation; key discriminative semantic features are leisure, work, and commuting periods. Our approach achieves high-accuracy stress recognition while enforcing strict location privacy.

Technology Category

• Machine Learning: Privacy
• Planning, Routing, and Scheduling: Activity and Plan Recognition
• Natural Language Processing: Safety and Robustness

Application Category

• Security and Privacy: Data transparency and provenance
• Semantics and Knowledge: Provenance, trust, security and privacy, and ethical issues in managing semantic data
• User Modeling, Personalization and Recommendation: User privacy protection in personalized systems

Abstract

Psychological stress is a widespread issue that significantly impacts student well-being and academic performance. Effective remote stress recognition is crucial, yet existing methods often rely on wearable devices or GPS-based clustering techniques that pose privacy risks. In this study, we introduce a novel, end-to-end privacy-enhanced framework for semantic location encoding using a self-hosted OSM engine and an LLM-bootstrapped static map. We rigorously quantify the privacy-utility trade-off and demonstrate (via LOSO validation) that our Privacy-Aware (PA) model achieves performance statistically indistinguishable from a non-private model, proving that utility does not require sacrificing privacy. Feature importance analysis highlights that recreational activity time, working time, and travel time play a significant role in stress recognition.

Problem

• Quantifying privacy-utility trade-offs in GPS stress recognition systems
• Developing privacy-enhanced semantic location encoding for stress detection
• Evaluating recreational and work activity patterns for stress recognition

Innovation

• Self-hosted OSM engine for semantic location encoding
• LLM-bootstrapped static map enhances privacy protection
• Privacy-aware model maintains utility without sacrificing privacy