Black-box Adversarial Attacks with Limited Queries and Information
This addresses security vulnerabilities in real-world AI systems by defining and testing more restrictive threat models, though it is incremental in refining attack methods for practical constraints.
The paper tackles the problem of adversarial attacks on neural network classifiers under realistic black-box threat models with limited queries and information, and demonstrates effective attacks against an ImageNet classifier and the Google Cloud Vision API.
Current neural network-based classifiers are susceptible to adversarial examples even in the black-box setting, where the attacker only has query access to the model. In practice, the threat model for real-world systems is often more restrictive than the typical black-box model where the adversary can observe the full output of the network on arbitrarily many chosen inputs. We define three realistic threat models that more accurately characterize many real-world classifiers: the query-limited setting, the partial-information setting, and the label-only setting. We develop new attacks that fool classifiers under these more restrictive threat models, where previous methods would be impractical or ineffective. We demonstrate that our methods are effective against an ImageNet classifier under our proposed threat models. We also demonstrate a targeted black-box attack against a commercial classifier, overcoming the challenges of limited query access, partial information, and other practical issues to break the Google Cloud Vision API.