Towards Privacy Preserving HAR in Contested Environments: An FHE-ORAM Architecture

Real-time military personnel activity detection can be achieved through wearable sensors for enhanced operational readiness and safety during training and on the modern battle-field. Data collected from wearable sensors can prevent personnel injuries and aid tactical decision-making. However, several privacy and security issues arise from collecting physiological data that can reveal sensitive information about the coordinates and fatigue level of the personnel, which could compromise operational security. In this paper, we present a novel system that enables secure and privacy-preserving inference on the public cloud infrastructure by leveraging the benefits of Fully Homomorphic Encryption (FHE) techniques for end-to-end encryption and Oblivious Random Access Memory (ORAM) for mitigating the leakage in memory access patterns. A thorough security analysis and experimental validation validate the feasibility of our proposed approach, achieving an 87.45% activity detection accuracy and paving the way forward for guaranteeing end-to-end privacy in untrusted operational environments.