MIBench: A Comprehensive AI Benchmark for Model Inversion Attack and Defense
A Model Inversion (MI) attack is a type of privacy attack on machine learning and deep learning models, where an attacker tries to invert the model’s outputs to recreate privacy-sensitive…