Revisiting Transferable Adversarial Images: Systemization, Evaluation, and New Insights
This work addresses security concerns for computer vision systems in real-world black-box scenarios by providing a systematic evaluation framework, though it is incremental in nature.
The paper tackles the lack of systematic evaluation in transferable adversarial image attacks by categorizing attacks and conducting a comprehensive assessment against various defenses, revealing that early attacks can outperform newer ones and that state-of-the-art defenses are vulnerable to transfer attacks.
Transferable adversarial images raise critical security concerns for computer vision systems in real-world, black-box attack scenarios. Although many transfer attacks have been proposed, existing research lacks a systematic and comprehensive evaluation. In this paper, we systemize transfer attacks into five categories around the general machine learning pipeline and provide the first comprehensive evaluation, with 23 representative attacks against 11 representative defenses, including the recent, transfer-oriented defense and the real-world Google Cloud Vision. In particular, we identify two main problems of existing evaluations: (1) for attack transferability, lack of intra-category analyses with fair hyperparameter settings, and (2) for attack stealthiness, lack of diverse measures. Our evaluation results validate that these problems have indeed caused misleading conclusions and missing points, and addressing them leads to new, \textit{consensus-challenging} insights, such as (1) an early attack, DI, even outperforms all similar follow-up ones, (2) the state-of-the-art (white-box) defense, DiffPure, is even vulnerable to (black-box) transfer attacks, and (3) even under the same $L_p$ constraint, different attacks yield dramatically different stealthiness results regarding diverse imperceptibility metrics, finer-grained measures, and a user study. We hope that our analyses will serve as guidance on properly evaluating transferable adversarial images and advance the design of attacks and defenses. Code is available at https://github.com/ZhengyuZhao/TransferAttackEval.