Prediction Poisoning: Utility-Constrained Defenses Against Model Stealing Attacks

Open in new window