We present Neo, a blackbox tool to detect backdoor attacks in Machine Learning models. Please see the paper Model Agnostic Defence against Backdoor Attacks in Machine Learning for more details.
We evaluate Neo against two attacks, BadNets and TrojanNN.
The MNIST and USTS datasets are part of BadNets whereas the VGG-Face dataset are part of TrojanNN.
- Please contact [email protected] for any comments/questions