Vax-a-Net: Training-time Defence Against Adversarial Patch Attacks
This work addresses security vulnerabilities in CNNs for applications like autonomous systems, though it is incremental as it builds on existing adversarial training methods.
The authors tackled the problem of adversarial patch attacks on convolutional neural networks by introducing Vax-a-Net, a conditional GAN that synthesizes patches to adapt pre-trained models, resulting in improved resilience against such attacks across multiple architectures.
We present Vax-a-Net; a technique for immunizing convolutional neural networks (CNNs) against adversarial patch attacks (APAs). APAs insert visually overt, local regions (patches) into an image to induce misclassification. We introduce a conditional Generative Adversarial Network (GAN) architecture that simultaneously learns to synthesise patches for use in APAs, whilst exploiting those attacks to adapt a pre-trained target CNN to reduce its susceptibility to them. This approach enables resilience against APAs to be conferred to pre-trained models, which would be impractical with conventional adversarial training due to the slow convergence of APA methods. We demonstrate transferability of this protection to defend against existing APAs, and show its efficacy across several contemporary CNN architectures.