On-Device AI for Secure Patient Health Monitoring

The increasing adoption of artificial intelligence (AI) in mobile health applications has introduced critical concerns regarding user privacy and data security. Privacy risks, particularly when health data is transmitted to cloud servers, present a major barrier to widespread adoption of AI-driven personal health monitoring systems. Regulatory frameworks such as HIPAA and GDPR outline strict requirements for securing patient data, yet achieving compliance in mobile environments remains challenging. Previous research has explored encryption techniques and secure data transmission methods to address privacy concerns. While effective to some extent, these methods still rely on cloud infrastructure, creating vulnerabilities and limiting offline functionality. Lightweight on-device AI models have emerged as a promising alternative, but ensuring privacy without compromising diagnostic accuracy presents a technical challenge. In this paper, we propose an on-device AI model that integrates TensorFlow Lite for lightweight inference and the Gemma 2B language model for personalized health insights. Our model addresses key privacy concerns by analyzing health data locally on the user’s device, ensuring secure, private assessments without relying on cloud connectivity. We implemented an on-device AI module for patient health monitoring as a proof of concept.