Securing AI and Agentic Systems in Medical Devices: Methods, Risks, and Defenses
Authors:
- 25 September 2025
- conference paper
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-8
Abstract
Integrating AI and agentic systems into medical devices significantly enhances healthcare by enabling autonomous diagnosis, personalized treatment, and real-time monitoring. However, this advancement introduces complex security challenges, including risks related to autonomous decision-making, multi-agent coordination, and AI-specific attacks that can jeopardize patient safety. This whitepaper outlines a framework for securing AI-enabled medical devices, focusing on critical vulnerabilities and risk prioritization, while proposing multi-layered defense strategies based on established cybersecurity principles. It systematically addresses security risks throughout the AI system lifecycle using FDA guidance, EU regulations like GDPR and the AI Act, and the MITRE ATLAS framework. The paper concludes with a case study on implementing this framework in an AWS healthcare AI system, showcasing its practical application.Keywords
This publication has 8 references indexed in Scilit:
- Evaluating accountability, transparency, and bias in AI-assisted healthcare decision- making: a qualitative study of healthcare professionals’ perspectives in the UKBMC Medical Ethics, 2025
- AI Agents Under Threat: A Survey of Key Security Challenges and Future PathwaysACM Computing Surveys, 2025
- Towards secure and trusted AI in healthcare: A systematic review of emerging innovations and ethical challengesInternational Journal of Medical Informatics, 2024
- Large language models encode clinical knowledgeNature, 2023
- Advances in Continuous Glucose Monitoring and Integrated Devices for Management of Diabetes with Insulin-Based Therapy: Improvement in Glycemic ControlDiabetes & Metabolism Journal, 2023
- Role of Artificial Intelligence in Patient Safety Outcomes: Systematic Literature ReviewJMIR Medical Informatics, 2020
- Cybersecurity data science: an overview from machine learning perspectiveJournal of Big Data, 2020
- Dermatologist-level classification of skin cancer with deep neural networksNature, 2017