Query-Efficient Black-box Attack

[sroshan89]

Hello,
I'm checking the available black-box evasion attacks in ART. In the Wiki, it's mentioned that the following attack is supported Query-Efficient Black-box Adversarial Examples. But I couldn't locate the implementation in the repository.
I'd be thankful if you could provide a pointer to the code.

[beat-buesser]

Hi @sroshan89 I think this black-box method has been implemented as a seprate estimator in https://github.com/Trusted-AI/adversarial-robustness-toolbox/blob/main/art/estimators/classification/query_efficient_bb.py. It provides black-box gradient estimations in method loss_gradients that can be used with all white-box attacks.