search

Trusted-AI / adversarial-robustness-toolbox

Adversarial Robustness Toolbox (ART) - Python Library for Machine Learning Security - Evasion, Poisoning, Extraction, Inference - Red and Blue Teams
https://adversarial-robustness-toolbox.readthedocs.io/en/latest/
MIT License
4.88k stars 1.17k forks source link

Deepfool attack not applied #1055

Closed aqsaimtiaz closed 3 years ago

aqsaimtiaz commented 3 years ago

I am trying to apply the whitebox attacks on digit images, the prediction for digits are accurate, but when the attack is applied, the losses is 0.0 and the init images remain the same and does not change into target image. colab notebook also attached for reference. Your help will be highly appreciated to debug this.

https://drive.google.com/file/d/1dfUBV45lfMumN6UI00w4KX4Bgid-b43l/view?usp=sharing

beat-buesser commented 3 years ago

Hi @aqsaimtiaz

One issue could be that your model is predicting probabilities whereas DeepFool is expecting logits as prediction output.