ML Security PapersA Practical Framework for Evaluating Medical AI Security: Reproducible Assessment of Jailbreaking and Privacy Vulnerabilities Across Clinical Specialties
Jinghao Wang , Ping Zhang , Carter Yagemann
Published on arXiv
2512.08185
Prompt Injection
OWASP LLM Top 10 — LLM01
Sensitive Information Disclosure
OWASP LLM Top 10 — LLM06
Key Finding
Framework specification (threat models, synthetic data methodology, evaluation protocols, scoring rubrics) enabling reproducible security benchmarking of medical LLMs without GPU clusters, commercial API access, or protected health data.
Medical Large Language Models (LLMs) are increasingly deployed for clinical decision support across diverse specialties, yet systematic evaluation of their robustness to adversarial misuse and privacy leakage remains inaccessible to most researchers. Existing security benchmarks require GPU clusters, commercial API access, or protected health data -- barriers that limit community participation in this critical research area. We propose a practical, fully reproducible framework for evaluating medical AI security under realistic resource constraints. Our framework design covers multiple medical specialties stratified by clinical risk -- from high-risk domains such as emergency medicine and psychiatry to general practice -- addressing jailbreaking attacks (role-playing, authority impersonation, multi-turn manipulation) and privacy extraction attacks. All evaluation utilizes synthetic patient records requiring no IRB approval. The framework is designed to run entirely on consumer CPU hardware using freely available models, eliminating cost barriers. We present the framework specification including threat models, data generation methodology, evaluation protocols, and scoring rubrics. This proposal establishes a foundation for comparative security assessment of medical-specialist models and defense mechanisms, advancing the broader goal of ensuring safe and trustworthy medical AI systems.
Key Contributions
- Multi-specialty threat model stratified by clinical risk (emergency medicine, psychiatry, general practice) covering both jailbreaking and privacy extraction attack scenarios
- Accessible evaluation design that runs on consumer CPU hardware using freely available models and synthetic patient records requiring no IRB approval
- Standardized evaluation protocol with scoring rubrics adapted from established security research for comparative assessment of medical LLM defenses