Differentially Private EEG Feature Anonymization: A Privacy-Utility Case Study in Clinical Neurophysiology
Case study applies Gaussian and Laplace differential privacy to clinical EEG features, quantifying privacy-utility trade-offs across three deployment scenarios.
Researchers evaluate subject-level differential privacy for EEG-derived feature representations using Gaussian and Laplace perturbations across client-side, centralized server-side, and decentralized local training scenarios. Utility is assessed with statistical measures and a downstream machine-learning check on clinical neurophysiology data. Results show DP can be integrated into EEG workflows, but mechanism choice, privacy parameters, and sensitivity calibration strongly influence data utility, particularly on small and imbalanced clinical datasets. The study highlights the privacy-utility trade-off in protecting biomedical signals against re-identification and inference risks.
- Covers client-side, centralized, and decentralized anonymization scenarios
- Gaussian and Laplace mechanisms compared on utility metrics
- Small, imbalanced clinical datasets amplify privacy-utility trade-offs
- Standard de-identification insufficient for high-dimensional biomedical signals
From the original article
1 min read at arxiv.orgClinical electroencephalography (EEG) data are valuable for healthcare research and for developing artificial intelligence (AI)-based clinical decision-support systems, but EEG recordings and derived features may contain sensitive patient-specific information. This creates privacy risks when data are reused, analyzed, or shared across clinical and research environments.
By Noman Sadiq · arxiv.org
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.