PromptLA: Towards Integrity Verification of Black-box Text-to-Image Diffusion Models

TL;DR

This paper introduces PromptLA, a novel method for verifying the integrity of black-box text-to-image diffusion models by analyzing feature distributions, achieving high accuracy and robustness while addressing regulatory and legal concerns.

Contribution

The paper presents the first approach for integrity verification of T2I diffusion models using a prompt selection algorithm and distribution analysis, improving detection accuracy and efficiency.

Findings

Achieves over 0.96 AUC in integrity detection

Outperforms baseline methods by more than 0.2 in AUC

Robust against image-level post-processing

Abstract

Despite the impressive synthesis quality of text-to-image (T2I) diffusion models, their black-box deployment poses significant regulatory challenges: Malicious actors can fine-tune these models to generate illegal content, circumventing existing safeguards through parameter manipulation. Therefore, it is essential to verify the integrity of T2I diffusion models. To this end, considering the randomness within the outputs of generative models and the high costs in interacting with them, we discern model tampering via the KL divergence between the distributions of the features of generated images. We propose a novel prompt selection algorithm based on learning automaton (PromptLA) for efficient and accurate verification. Evaluations on four advanced T2I models (e.g., SDXL, FLUX.1) demonstrate that our method achieves a mean AUC of over 0.96 in integrity detection, exceeding baselines by more than 0.2, showcasing strong effectiveness and generalization. Additionally, our approach achieves lower cost and is robust against image-level post-processing. To the best of our knowledge, this paper is the first work addressing the integrity verification of T2I diffusion models, which establishes quantifiable standards for AI copyright litigation in practice.