UMass AI Security

Our Expertise

🎨

Audio, text-to-image, reasoning LLMs and beyond — securing AI across all the ways it sees and creates

Text-to-Image Audio Reasoning LLM

Current Research: Text-to-Image Generation Security, Audio LLM Security, and Reasoning Time Attacks

🔒

Membership inference attacks, contextual integrity, and keeping your data yours in the age of AI

Membership Inference Memorization Contextual Integrity

Current Research: Memorization, Data Leakage, Privacy-Preserving AI systems, and Contextual Integrity frameworks

🕷️

Controlling and securing autonomous agents across all modalities

Web Agent Research Agent Multi-Modal Agent

Current Research: Reasoning Gates for Web Agents, Controlling AI Research Agents, and Securing Multi-Modal Agents

⚖️

Understanding and building AI systems that align with human values and societal norms

Bias Propaganda Responsible AI

Current Research: Bias Attack and Detection in Text-to-Image, Fairness in LLMs