Adversarial example soups: averaging multiple adversarial examples improves transferability without increasing additional generation time

arXiv:2402.18370v1 Announce Type: new
Abstract: For transfer-based attacks, the adversarial examples are crafted on the surrogate model, which can be implemented to mislead the target model effectively. The conventional method for maximizing adversarial transferability involves: (1) fine-tuning hyperparameters to generate multiple batches of adversarial examples on the substitute model; (2) conserving the batch of adversarial examples that have the best comprehensive performance on substitute model and target model, and discarding the others. In this work, we revisit the second step …

abstract adversarial adversarial examples arxiv attacks cs.cv example examples fine-tuning generate multiple transfer type