#membership-inference

Membership Inference Attacks: What Actually Works Against Production ML APIs

Shokri et al.'s shadow-model attack is the canonical reference, but the gap between the paper's threat model and a real rate-limited API is wide. Here's what survives that gap.