Spy-Watermark: Robust Invisible Watermarking for Backdoor Attack

TL;DR

Spy-Watermark introduces a robust, invisible watermark-based backdoor attack that remains effective under data corruption and defenses, outperforming existing methods in robustness and stealthiness across multiple datasets.

Contribution

The paper presents a novel latent domain watermarking technique for backdoor attacks that is highly resistant to data collapse and defenses, improving robustness and stealthiness.

Findings

Outperforms ten state-of-the-art methods in robustness.

Effective across CIFAR10, GTSRB, and ImageNet datasets.

Demonstrates high resilience against data corruption and defenses.

Abstract

Backdoor attack aims to deceive a victim model when facing backdoor instances while maintaining its performance on benign data. Current methods use manual patterns or special perturbations as triggers, while they often overlook the robustness against data corruption, making backdoor attacks easy to defend in practice. To address this issue, we propose a novel backdoor attack method named Spy-Watermark, which remains effective when facing data collapse and backdoor defense. Therein, we introduce a learnable watermark embedded in the latent domain of images, serving as the trigger. Then, we search for a watermark that can withstand collapse during image decoding, cooperating with several anti-collapse operations to further enhance the resilience of our trigger against data corruption. Extensive experiments are conducted on CIFAR10, GTSRB, and ImageNet datasets, demonstrating that Spy-Watermark overtakes ten state-of-the-art methods in terms of robustness and stealthiness.