LoMime

New LoMime method enables efficient privacy attacks on ML models

Researchers have developed a novel method called LoMime for performing membership inference attacks (MIAs) on machine learning models, even when only label-only access is available. This approach leverages model extract…