Lilith

New 'Lilith' framework enables backdoor generalization in ML models

Researchers have introduced Lilith, a new framework designed to address backdoor generalization in machine learning models. This framework focuses on implanting malicious behavior that can generalize to inference-time t…