Implementation of badnets on CIFAR10 and MNIST
This is a pytorch implementation of the paper BadNets: Identifying Vulnerabilities in the Machine Learning Model Supply Chain, https://arxiv.org/abs/1708.06733
A backdoor attack is developed, by poisoning part of the dataset with a white pixel pattern as shown in the paper. Both triggers, ie single target attack, as well as, the all-to-all attack, are developed as described on the paper.