adversarial-robustness-toolbox
adversarial-robustness-toolbox copied to clipboard
add L-inf bound to `SimBA`
Is your feature request related to a problem? Please describe.
add the option to use an L-infinity bound with the SimBA
attack
Describe the solution you'd like
adapt for art
the approach recently added by the authors to their implementation:
https://github.com/cg563/simple-blackbox-attack/blob/418b25f4dd8cc0f988376c5730ae3988b95fbce0/simba.py#L58
Describe alternatives you've considered No obvious alternatives I can think of
Additional context as mentioned in #1410 -- and relevant here, this will probably require adding a unit test or two