Attention to Neural Plagiarism: Diffusion Models Can Plagiarize Your Copyrighted Images!

TL;DR

This paper reveals that diffusion neural models can plagiarize copyrighted images by bypassing watermark protections using a gradient-based method, highlighting a significant threat to intellectual property.

Contribution

It introduces a novel gradient-based approach using anchors and shims to replicate or create ambiguity in copyrighted images without additional training.

Findings

Diffusion models can replicate copyrighted images.

Watermark protections can be bypassed using the proposed method.

The approach requires no extra training or fine-tuning.

Abstract

In this paper, we highlight a critical threat posed by emerging neural models: data plagiarism. We demonstrate how modern neural models (e.g., diffusion models) can replicate copyrighted images, even when protected by advanced watermarking techniques. To expose vulnerabilities in copyright protection and facilitate future research, we propose a general approach to neural plagiarism that can either forge replicas of copyrighted data or introduce copyright ambiguity. Our method, based on "anchors and shims", employs inverse latents as anchors and finds shim perturbations that gradually deviate the anchor latents, thereby evading watermark or copyright detection. By applying perturbations to the cross-attention mechanism at different timesteps, our approach induces varying degrees of semantic modification in copyrighted images, enabling it to bypass protections ranging from visible trademarks and signatures to invisible watermarks. Notably, our method is a purely gradient-based search that requires no additional training or fine-tuning. Experiments on MS-COCO and real-world copyrighted images show that diffusion models can replicate copyrighted images, underscoring the urgent need for countermeasures against neural plagiarism.