PRADA: Protecting against DNN Model Stealing Attacks
This addresses the security issue of protecting proprietary ML models from theft for businesses and researchers, representing an incremental advance in detection and attack methods.
The paper tackles the problem of DNN model stealing via prediction APIs by developing new model extraction attacks that outperform state-of-the-art methods, achieving up to +44 percentage points in adversarial example transferability and +46 percentage points in prediction accuracy, and proposes PRADA, a detection method that detects all prior attacks with no false positives.
Machine learning (ML) applications are increasingly prevalent. Protecting the confidentiality of ML models becomes paramount for two reasons: (a) a model can be a business advantage to its owner, and (b) an adversary may use a stolen model to find transferable adversarial examples that can evade classification by the original model. Access to the model can be restricted to be only via well-defined prediction APIs. Nevertheless, prediction APIs still provide enough information to allow an adversary to mount model extraction attacks by sending repeated queries via the prediction API. In this paper, we describe new model extraction attacks using novel approaches for generating synthetic queries, and optimizing training hyperparameters. Our attacks outperform state-of-the-art model extraction in terms of transferability of both targeted and non-targeted adversarial examples (up to +29-44 percentage points, pp), and prediction accuracy (up to +46 pp) on two datasets. We provide take-aways on how to perform effective model extraction attacks. We then propose PRADA, the first step towards generic and effective detection of DNN model extraction attacks. It analyzes the distribution of consecutive API queries and raises an alarm when this distribution deviates from benign behavior. We show that PRADA can detect all prior model extraction attacks with no false positives.