Revisiting Transferable Adversarial Image Examples: Attack Categorization, Evaluation Guidelines, and New Insights. (arXiv:2310.11850v1 [cs.CR])

Transferable adversarial examples raise critical security concerns in
real-world, black-box attack scenarios. However, in this work, we identify two
main problems in common evaluation practices: (1) For attack transferability,
lack of systematic, one-to-one attack comparison and fair hyperparameter
settings. (2) For attack stealthiness, simply no comparisons. To address these
problems, we establish new evaluation guidelines by (1) proposing a novel
attack categorization strategy and conducting systematic and fair
intra-category analyses on transferability, and (2) considering diverse
imperceptibility metrics and finer-grained …

adversarial attack box comparisons critical evaluation fair guidelines identify image insights main practices problems security security concerns settings work world