Vitaly Shmatikov
Cornell University
Time: Friday, Apr. 4 12:30 PM - 1:30 PM Location: MKB 622
What You See Is Not What You Get: Multi-Modal AI Systems Are Not Secure
Abstract:
Modern AI/ML systems (in particular, semantic retrieval and LLM-based systems) accept not just text inputs but also images, audio, video, and other modalities. In this talk, I will show how attackers can exploit non-text vectors for spamming, misinformation, malicious code execution, and other adversarial objectives. I will also discuss why adversarial robustness seems difficult to achieve in multi-modal systems.
Bio:
Vitaly Shmatikov is a Professor of Computer Science at Cornell University and Cornell Tech. Research by Dr. Shmatikov, his students, and collaborators received the Caspar Bowden PET Award for Outstanding Research in Privacy Enhancing Technologies three times; Test-of-Time Awards from the IEEE Symposium on Security and Privacy (S&P / “Oakland”), ACM Conference on Computer and Communications Security (CCS), and the ACM/IEEE Symposium on Logic in Computer Science (LICS); as well as several outstanding and distinguished paper awards, most recently from USENIX Security 2021 and 2024 and EMNLP 2023.
Charlie Epperson
U.S. Coast Guard
Time: Friday, Mar. 7 12:30 PM - 1:30 PM Location: MKB 622
Bridging the Gap: AI R&D and Real-World Maritime Intelligence Application
Abstract:
The intersection of maritime intelligence and artificial intelligence presents transformative opportunities for enhancing maritime domain awareness. This presentation explores how the Coast Guard and partners have pursued AI capabilities, including computer vision, generative AI, and autonomous systems, to address critical maritime challenges such as counter-narcotics, migration, and search and rescue. CDR Epperson will detail ongoing interagency initiatives focused on developing organic AI capabilities and integrating cutting-edge commercial solutions. Furthermore, he will outline key engagement pathways for university AI researchers to contribute to the U.S. Government's AI efforts.
Bio:
Commander Charlie Epperson serves as the Deputy Chief of Artificial Intelligence within the U.S. Coast Guard's Office of Intelligence, where he leads efforts to integrate advanced AI technologies into maritime operations. Prior to this role, he was the Acting Director of Humanitarian Assistance and Disaster Relief (HADR) at the Department of Defense's Joint Artificial Intelligence Center (JAIC), spearheading initiatives to modernize disaster response through AI, including wildfire mitigation, rapid damage assessment, and enhanced search and rescue. With 22 years of operational experience, CDR Epperson has served in diverse and challenging environments, including a five-year assignment in Guam focusing on search and rescue, and deployments with the National Strike Force responding to the BP Deepwater Horizon oil spill and major hurricane response operations. CDR Epperson holds a Master of Public Administration (MPA) from the Lee Kuan Yew School of Public Policy at the National University of Singapore, a graduate certificate in Community Preparedness & Disaster Management (CPDM) from the University of North Carolina and completed the National Preparedness Leadership Initiative at Harvard University. He is a proud alumnus and Letterman of the University of Tennessee. Outside of his professional life, CDR Epperson and his wife are avid long-distance runners and dedicated Tennessee baseball fans.