Provably Unlearnable Data Examples
Derui Wang
Network and Distributed System Security (NDSS) Symposium 2025 · Day 3 · Machine Unlearning
In an era dominated by large language models and advanced machine learning, the ease with which public data can be exploited poses significant risks, ranging from intellectual property infringement to privacy breaches. Derui Wang's talk, "Provably Unlearnable Data Examples," introduces a groundbreaking framework designed to certify the learnability of data, offering robust protection against unauthorized model training and exploitation. This work, a collaborative effort from CSRO's Data61 and the University of Chicago, and supported by the Cyber Security Cooperative Research Centre of Australia, addresses the critical need for a quantifiable guarantee on data's unlearnability.
AI review
Solid, technically grounded ML security research that fills a real gap: the first certification framework for unlearnable data examples, complete with provable upper bounds on adversarial utility extraction. The recovery attack identification alone justifies the slot — it's a novel threat vector that invalidates prior defenses and the proposed RWP-based countermeasure has clear theoretical backing.