Enhancing Targeted Attack Transferability via Diversified Weight Pruning
This work addresses a security vulnerability in neural networks for malicious attackers, offering an incremental improvement over existing methods by extending model augmentation to targeted attacks.
The paper tackles the problem of generating targeted adversarial examples that transfer across models by proposing Diversified Weight Pruning (DWP), a model augmentation technique that improves targeted attack success rates by up to 10.1%, 6.6%, and 7.0% in various challenging scenarios.
Malicious attackers can generate targeted adversarial examples by imposing tiny noises, forcing neural networks to produce specific incorrect outputs. With cross-model transferability, network models remain vulnerable even in black-box settings. Recent studies have shown the effectiveness of ensemble-based methods in generating transferable adversarial examples. To further enhance transferability, model augmentation methods aim to produce more networks participating in the ensemble. However, existing model augmentation methods are only proven effective in untargeted attacks. In this work, we propose Diversified Weight Pruning (DWP), a novel model augmentation technique for generating transferable targeted attacks. DWP leverages the weight pruning method commonly used in model compression. Compared with prior work, DWP protects necessary connections and ensures the diversity of the pruned models simultaneously, which we show are crucial for targeted transferability. Experiments on the ImageNet-compatible dataset under various and more challenging scenarios confirm the effectiveness: transferring to adversarially trained models, Non-CNN architectures, and Google Cloud Vision. The results show that our proposed DWP improves the targeted attack success rates with up to $10.1$%, $6.6$%, and $7.0$% on the combination of state-of-the-art methods, respectively. The source code will be made available after acceptance.