Invited Talk 1: Reza Shokri - On Testing Memorization in AI: From Brute-Force Methods to Robust Statistical Tests
2025 Invited Talk
in
Workshop: The Impact of Memorization on Trustworthy Foundation Models
in
Workshop: The Impact of Memorization on Trustworthy Foundation Models
Abstract
In this talk, the speaker will explore a range of techniques for detecting and analyzing memorization in machine learning models. The presentation will cover proof-of-concept brute-force extraction attacks, high-power and range-based membership inference attacks, as well as recent advances in dataset inference methods. The strengths and limitations of each approach will be discussed, offering a comprehensive perspective on the current landscape of memorization testing.
Video
Chat is not available.
Successful Page Load