Dongxian Wu

backdoor attack (2) adversarial training (2) adversarial robustness (2) backdoor defense (1) model architecture (1) loss landscape (1) deep neural network (1) adversarial defense (1) ownership verification (1) removal attack (1) pretrained model (1) robust generalization (1) model purification (1) model watermarking (1) neural network (1) adversarial weight perturbation (1) gradient masking (1) backdoor removal (1) parametric vulnerability (1) neural network pruning (1)