adversarial-robustness-toolbox
adversarial-robustness-toolbox copied to clipboard
Trusted-AI
Adversarial Robustness Toolbox (ART) - Python Library for Machine Learning Security - Evasion, Poisoning, Extraction, Inference - Red and Blue Teams
**Describe the bug** Using the `ImperceptibleASRPyTorch` normally may not cause this problem, NAN loss, since similar issues have been fixed. But when I use the `self._psd_transform`, `self._compute_masking_threshold` and `self._forward_2nd_stage` functions...
To support better dynamic mask attacks as well as add physical attack support, we should integrate the GRAPHITE attack (https://arxiv.org/abs/2002.07088), which supports blackbox and whitebox attacks both in digital and...
**Describe the bug** There are some errors while running the notebook in https://github.com/Trusted-AI/adversarial-robustness-toolbox/blob/main/notebooks/expectation_over_transformation_classification_rotation.ipynb The first one is that the following part is wrong ``` from art.classifiers import TensorFlowV2Classifier ``` It...
We would like to investigate and add support for multiple GPUs for PyTorch estimators.
We would like to extend support for Indicators of Attack Failure to include speech recognition and object detection/tracking. To facilitate this extension we would like to refactor implementation by introducing...
**Describe the bug** Current implementation of the Bullseye polytope attack does not reproduce the expected behavior: attack doesn't work. **Expected behavior** Ensure the code works for non-sequential models. I recommend...
**Is your feature request related to a problem? Please describe.** Currently the docstring for `Simba.generate` states that `y` should be class labels (as is true for most `art` attacks). https://github.com/Trusted-AI/adversarial-robustness-toolbox/blob/85ca10ff6b696dbaa66dcf179d0493b12ff594bb/art/attacks/evasion/simba.py#L104...
**Is your feature request related to a problem? Please describe.** Boosting Adversarial Attacks with Momentum https://arxiv.org/pdf/1710.06081
**Is your feature request related to a problem? Please describe.** What Doesn’t Kill You Makes You Robust(er): Adversarial Training against Poisons and Backdoors https://arxiv.org/pdf/2102.13624.pdf **Describe the solution you'd like** -...
**Is your feature request related to a problem? Please describe.** What Doesn’t Kill You Makes You Robust(er): Adversarial Training against Poisons and Backdoors https://arxiv.org/pdf/2102.13624.pdf **Describe the solution you'd like** -...
Metadata
Owner
Metadata
Adversarial Robustness Toolbox (ART) - Python Library for Machine Learning Security - Evasion, Poisoning, Extraction, Inference - Red and Blue Teams