UMass AI Security Lab

Faculty

Amir Houmansadr

Amir Houmansadr

Professor

 Eugene Bagdasarian

Eugene Bagdasarian

Assistant Professor

 Shiqing Ma

Shiqing Ma

Assistant Professor (Affiliated)

 Adam O'Neill

Adam O'Neill

Assistant Professor (Affiliated)

Recent News

COLM 2025
New Blog Post
Thoughts on COLM 2025
Jaechul Roh
IEEE S&P 2026
IEEE S&P 2026
Exploiting Leaderboards for Large-Scale Distribution of Malicious Models
Anshuman Suri, Harsh Chaudhari, Yuefeng Peng, Ali Naseh, Amir Houmansadr, Alina Oprea
Terrarium
New Demo
Terrarium: Multi-Agent Safety, Privacy, and Security Studies
Mason Nakamura, Abhinav Kumar, Saaduddin Mahmud, Sahar Abdelnabi, Shlomo Zilberstein, Eugene Bagdasarian

Research Areas

🎨

Attacks on Generative Models

Securing AI across all modalities — audio, text-to-image, reasoning LLMs and beyond

Text-to-Image Audio Reasoning LLM
Current Focus: Text-to-Image Security, Audio LLM Security, Reasoning Time Attacks
🔒

AI Privacy

Membership inference, contextual integrity, and protecting data in the age of AI

Membership Inference Memorization Contextual Integrity
Current Focus: Data Leakage, Privacy-Preserving AI, Contextual Integrity Frameworks
🤖

Agentic Security

Controlling and securing autonomous agents across all modalities

Web Agents Research Agents Multi-Modal
Current Focus: Reasoning Gates for Web Agents, AI Research Agent Control
⚖️

Ethics & Societal Impact

Building AI systems that align with human values and societal norms

Bias Detection Propaganda Responsible AI
Current Focus: Bias in Text-to-Image, Fairness in LLMs