carlini / nn_robust_attacks

Robust evasion attacks against neural network to find adversarial examples

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Why the label of attack images is same with the original images?

shenqixiaojiang opened this issue · comments

The label of attack images is same with the original images when I use the run_inference_on_image function of setup_inception file.
So does it mean that the attack almost failed?

Sorry I don't quite understand what you're asking. If the prediction is the same for both before and after attack, then the attack has failed. Sometimes this might be because confidence=0 has minor issues, and if you run it with confidence=[some small value] it will often work.

@carlini Thanks,I will try the attack with confidence > 0.