Safeguarding the Artificial Pancreas: A Review of Security and Reliability Gaps and AI Driven Resilience

Rapid adoption of artificial pancreatic systems (AP) for the management of type 1 diabetes (T1D) requires a critical examination of their cybersecurity vulnerabilities, particularly in light of emerging AI-driven threats. This systematization of knowledge paper provides a comprehensive overview of AP systems, analyzing their current state, identifying critical vulnerabilities across software, hardware, and communication components, and exploring the dual role of artificial intelligence (AI). We show that while AI techniques, such as reinforcement learning, deep learning, and federated learning, offer promising solutions to improve security, performance, and personalization, they also introduce new attack vectors, including adversarial attacks on control algorithms and AI-enhanced exploitation of existing vulnerabilities. This paper systematically categorizes known and potential attacks, synthesizes current research on AI-driven threats and defenses, and identifies critical research gaps that must be addressed to ensure patient safety. We highlight the urgent need for robust, explainable AI, multi-layered security approaches, and proactive collaboration between researchers, clinicians, manufacturers, and regulatory bodies to realize the full potential of secure and intelligent AP systems.