Organizations: Laboratory of Intelligent Collaborative Computing, University of Electronic Science and Technology of China, China · School of Computer Science and Engineering, University of Electronic Science and Technology of China, China
With the rise of diverse Text-to-Image (T2I) models such as DALL-E 3, third-party platforms increasingly integrate multiple black-box APIs to deliver cost-effective services. This also raises a new and practical security question: are these third-party platforms truly offering the models they claim? Such model substitution fraud, in which a malicious third-party platform operator acts as the adversary, enables the platform to profit illicitly while undermining user interests and damaging the commercial reputation of legitimate model providers. To address this practical concern, we first define the task of T2I model verification, which determines whether a black-box target model matches a black-box oracle reference model. After that, we propose VerifyPrompt, which performs T2I model verification through a specially designed verify prompt. Intuitively, the verify prompt is an adversarial prompt for the target model, without transferability to other models. It makes the target model generate a specific image while making other models produce entirely different images. Our key idea for generating verify prompts is to enforce strong interactions between adversarial perturbations and the target model, thereby overfitting the target and producing model-specific perturbations. To enable efficient black-box search for verify prompts, we further introduce VerifyPrompt-RL, a reinforcement learning-based strategy that leverages multi-sample exploration and group relative error estimation to stabilize and accelerate the search process. Experiments show that VerifyPrompt achieves over 90% accuracy across various T2I platforms, confirming its effectiveness in real-world third-party services.
Figures & tables
Fig. 1 : Security issues in black-box API scenarios. The user requests DALL-E 3 from a third-party platform, which displays DALL-E 3 as the provided model on the interface but actually calls Stable Diffusion v1.4 behind the API.
Fig. 2 : Examples of whether adversarial prompts exhibit transferability. Adversarial prompts generated by RIATIG [ 5 ] , SneakyPrompt [ 6 ] and VerifyPrompt. The red text represents adversarial perturbations. RIATIG and SneakyPrompt not only successfully attacked the target model but also transferred to other models, whereas VerifyPrompt was only effective on the target model.
Fig. 3 : Verifier’s capability for verification. The verifier has black-box access to both the reference model and the target model to be verified.
Fig. 4 : The negative correlation between the transfer utility and the interaction. The correlation is computed using the Pearson correlation on the normalized interactions. The blue shade represents the 95% confidence interval of the regression.
Fig. 5 : The workflow of VerifyPrompt.
Fig. 6 : Generated images of verify prompts under different T2I models
Fig. 7 : Impact of query budget
Fig. 8 : Impact of prompt perturbation length
Reference Model
Target Model
R-1
R-3
Stable Diffusion v2.1
Stable Diffusion v1.4 [ 21 ]
0/10
1/10
Hunyuan DiT v1.2 [ 23 ]
0/10
0/10
Playground-v2.5 [ 41 ]
0/10
0/10
Stable Diffusion v3 [ 44 ]
0/10
0/10
Hyper-SD [ 45 ]
0/10
0/10
SDXL [ 39 ]
0/10
0/10
TABLE IV : Model verification performance of VerifyPrompt on Hugging Face APIs.
Fig. 9 : Visualization results of our verify prompt for model APIs from Hugging Face
Fig. 10 : Visualization results of our verify prompt of DALL-E 3 from different APIs
Reference Model
Target Model
R-1
R-3
DALL-E 3 [ 3 ]
DALL-E 3
9/10
10/10
TABLE V : Model verification performance of VerifyPrompt on One-API APIs.
Fig. 11 : Semantic shift between original and adversarial prompts using DALL-E 3 as reference model.
Fine-tuning Steps
Verification Accuracy (%)
0
0.00
3000
72.60
5000
86.32
7000
88.30
9000
85.26
10000
93.26
TABLE VI : Verification accuracy at different fine-tuning steps on Stable Diffusion v2.1
Acc.
Prec.
Recall
F1
95.00
90.91
100.00
95.24
TABLE VII : Detection performance of the PPL-based detector.
Fig. 12 : PPL distributions of original prompts and VerifyPrompts.
As Text-to-Image (T2I) generation becomes widespread, third-party platforms increasingly integrate multiple model APIs for convenient image creation. However, false claims of using official models can mislead users and harm model owners' reputations, making model verification essential to confirm whether an API's underlying model matches its claim. Existing methods address this by using verification prompts generated by official model owners, but the generation relies on multiple reference models for optimization, leading to high computational cost and sensitivity to model selection. To address this problem, we propose a reference-free T2I model verification method called Boundary-aware Prompt Optimization (BPO). It directly explores the intrinsic characteristics of the target model. The key insight is that although different T2I models produce similar outputs for normal prompts, their semantic boundaries in the embedding space (transition zones between two concepts such as "corgi" and "bagel") are distinct. Prompts near these boundaries generate unstable outputs (e.g., sometimes a corgi and sometimes a bagel) on the target model but remain stable on other models. By identifying such boundary-adjacent prompts, BPO captures model-specific behaviors that serve as reliable verification cues for distinguishing T2I models. Experiments on five T2I models and four baselines demonstrate that BPO achieves superior verification accuracy.
Zidong Zhao, Yihao Huang, Qing Guo +5
Zhejiang University, Hangzhou, China · East China Normal University, Shanghai, China · Nankai University, Tianjin, China +4
Red-teaming Text-to-Image (T2I) models is essential for safe deployment, yet it remains particularly challenging against implicit adversarial prompts. Unlike explicit adversarial prompts that can be readily identified and blocked, implicit ones are much harder to detect: the prompts appear benign on the text surface yet still lead to inappropriate visual content. To address this, we propose Adversarial Probing for Implicit VulnErabilities (AdvPIE), a multimodal agentic framework to expose implicit vulnerabilities without requiring access to the parameters of target models. AdvPIE adopts a policy agent to generate and refine implicit adversarial prompts based on the feedback from a judge agent. To construct informative feedback, the judge agent provides modality-specific safety evaluation at both global and relative levels across iterations. To effectively leverage the feedback, we propose a novel Cumulative Adversarial Decoding strategy for the policy agent, which dynamically reweights token distributions to favor tokens that lead to more harmful images while preserving sampling diversity. Extensive experiments on standard and safety-aligned T2I models show that AdvPIE1 effectively uncovers implicit vulnerabilities, outperforming various baseline methods.
Chang Ma, Junlin Han, Shuo Chen +3
Columbia University, USA · University of Oxford, UK · LMU Munich, Germany +2
This paper investigates the challenging task of detecting backdoored text-to-image models under black-box settings and introduces a novel detection framework BlackMirror. Existing approaches typically rely on analyzing image-level similarity, under the assumption that backdoor-triggered generations exhibit strong consistency across samples. However, they struggle to generalize to recently emerging backdoor attacks, where backdoored generations can appear visually diverse. BlackMirror is motivated by an observation: across backdoor attacks, {only partial semantic patterns within the generated image are steadily manipulated, while the rest of the content remains diverse or benign. Accordingly, BlackMirror consists of two components: MirrorMatch, which aligns visual patterns with the corresponding instructions to detect semantic deviations; and MirrorVerify, which evaluates the stability of these deviations across varied prompts to distinguish true backdoor behavior from benign responses. BlackMirror is a general, training-free framework that can be deployed as a plug-and-play module in Model-as-a-Service (MaaS) applications. Comprehensive experiments demonstrate that BlackMirror achieves accurate detection across a wide range of attacks. Code is available at https://github.com/Ferry-Li/BlackMirror.
Feiran Li, Qianqian Xu, Shilong Bao +4
Institute of Information Engineering, CAS · School of Cyber Security, University of Chinese Academy of Sciences · State Key Laboratory of AI Safety, Institute of Computing Technology, CAS +4