1 paper · 1 filter
Minlong Peng, Zidi Xiong, Quang H. Nguyen +3
By injecting a small number of poisoned samples into the training set, backdoor attacks aim to make the victim model produce designed outputs on any input injected with pre-designe…