DSE-MSU / DeepRobust

A pytorch adversarial library for attack and defense methods on images and graphs
MIT License
976 stars 189 forks source link

Features Attack #123

Open VaporYuJ opened 1 year ago

VaporYuJ commented 1 year ago

Hi ~I wonder whether the global attacks in the deeprobust include featuresattacks , because I have not found the relevant implementation`~

VaporYuJ commented 1 year ago

Hi I wonder whether the global attacks in the deeprobust include feature attacks , because I have not found the relevant implementation`

ChandlerBang commented 1 year ago

Hey, you can refer to [the code in AirGNN]((https://github.com/lxiaorui/AirGNN/blob/main/adv_attack.py), which shows how to use deeprobust.graph.targeted_attack.Nettack for attacking features.

VaporYuJ commented 1 year ago

Hey, you can refer to [the code in AirGNN]((https://github.com/lxiaorui/AirGNN/blob/main/adv_attack.py), which shows how to use deeprobust.graph.targeted_attack.Nettack for attacking features.

Thanks~

VaporYuJ commented 1 year ago

hi~ Is there any solutions proposed on the global attack?Nettack is a kind of target....

ChandlerBang commented 1 year ago

I think you can apply PGD attack to generate features. But currently we do not support this function.

VaporYuJ commented 1 year ago

OK I will try THanks!