I will briefly discuss our demo system recently accepted by CCS 2024 that estimates privacy risks of participating in a machine learning task. As many AI models depend on sensitive training data (e.g., identifiable images and text conversations), privacy has increasingly been a major concern in the AI era. Methods like differential privacy allow users to quantify the acceptable privacy loss, which may also lead to significant utility loss. As a result, controlled data access is still the mainstream method for protecting data privacy or controlling privacy leakage in industrial and research settings.  However, there is no quantitative measure for individual data contributors to tell their privacy risks before participating in a machine learning task. We developed the demo prototype FT-PrivacyScore to efficiently and quantitatively estimate the privacy risk of participating in a model fine-tuning task. With FT-PrivacyScore, participants and data consumers can prepare proactively to ensure better protected privacy.

Dr. Keke Chen is an associate professor in the CSEE Department at UMBC. His recent research focuses on privacy and security issues with AI model training and deployment. He earned his Ph.D. in computer science from Georgia Tech in 2006. Before joining UMBC, he was a Northwestern Mutual associate professor of computer science at Marquette University. Email: kekechen@umbc.edu

Host: Alan T. Sherman, sherman@umbc.edu. Support for this event was provided in part by the NSF under SFS grant DGE-1753681. The UMBC Cyber Defense Lab meets biweekly Fridays 12-1pm. All meetings are open to the public. Upcoming CDL meetings: Nov. 15, Houbing Song (IS, UMBC) Dec. 6, Zhiuan Chen (IS, UMBC)