Welcome!

Join us in our Spring 2025 seminar COMPSCI 692PA on AI Security and Privacy.

New capabilities of language and diffusion models enable applications that interact with users across different modalities, perform independent actions, and leverage external tools. In this seminar, we will invite leading experts to discuss cutting-edge research in advancing the safety and trustworthiness of GenAI. This is a 1-credit class where students are expected to attend talks in person, participate in discussions, and write 2 thorough critiques of the speaker's presented work. Please join our Slack channel to stay up-to-date with the latest news and events.

Schedule

• The seminar will take place on Monday from 11am to 12pm ET (there may be some exceptions, see specific date below).
• You can attend virtually via Zoom or in person at CS 142.
• The schedule below is still being updated and maybe subject to changes, so check back occasionally.
• If you are interested in giving a seminar talk, please email or send us a Slack message (our contact info can be found here).

• Week 01: Mon, Feb 03 Anshuman Suri (Northeastern University): White-box vs Black-box: Privacy Auditing for Machine Learning (more)
• Week 02: Mon, Feb 10 Sahar Abdelnabi (Microsoft): Evaluating and Securing LLM-Agentic Networks (more)
• Week 03: Thu, Feb 20 Javier Rando (ETH Zurich): Gradient-based Jailbreak Images for Multimodal Fusion Models (more)
• Week 04: Mon, Feb 24 Norman Mu (xAI): A Closer Look at System Prompt Robustness (more)
• Week 05: Mon, Mar 03 Harsh Chaudhari (Northeastern University): Propagation of Adversarial Bias to Distilled Language Models (more)
• Week 06: Mon, Mar 10 Andy Zou (Carnegie Mellon University): Red Teaming AI Agents in-the-wild: Revealing Deployment Vulnerabilities (more)
• Week 07: Mon, Mar 24 Xiangyu Qi (OpenAI): Safety Alignment Should Be Made More Than Just A Few Tokens Deep
• Week 08: Mon, Mar 31 Om Thakkar (OpenAI): Privacy Leakage in Speech Models: Attacks and Mitigations
• Week 09: Mon, Apr 07 Ryan McKenna (Google): Differentially Private Machine Learning at Scale
• Week 10: Mon, Apr 14 Milad Nasr (Google): Topic TBD
• Week 11: Fri, Apr 18 Edoardo Debenedetti (ETH Zurich): Defeating Prompt Injections by Design
• Week 12: Mon, Apr 28 Jonas Geiping (ELLIS Institute): Topic TBD
• Week 13: Mon, May 05 Speaker TBD: Topic TBD