# CBV: Clean-label Backdoor Attacks on Vision Language Models via Diffusion Models

arXiv: https://arxiv.org/abs/2605.02202 · HTML: https://mlsanity.com/paper/2605.02202/

## Key facts

- Cited by 3 papers in the map (data through 14 Oct 2026).
- Placed in MLLM jailbreaks and adversarial robustness, one of 16 directions in Multimodal LLMs / vision-language models.
- Builds on 36 of its 48 references that are in the map.
- First posted on arXiv 4 May 2026.

## Placed in

- Multimodal LLMs / vision-language models / [MLLM jailbreaks and adversarial robustness](https://mlsanity.com/mllm/mllm-jailbreaks-and-adversarial-robustness/)
- Diffusion models / [Robustness, privacy and adversarial aspects](https://mlsanity.com/diffusion/robustness-privacy-and-adversarial-aspects/)
- LLM safety and security / [Backdoor and data poisoning attacks](https://mlsanity.com/safety/backdoor-and-data-poisoning-attacks/)

## Builds on (most cited first, 30 of 36)

- [Learning Transferable Visual Models From Natural Language Supervision](https://mlsanity.com/paper/2103.00020/) (2021)
- [High-Resolution Image Synthesis with Latent Diffusion Models](https://mlsanity.com/paper/2112.10752/) (2021)
- [LLaMA: Open and Efficient Foundation Language Models](https://mlsanity.com/paper/2302.13971/) (2023)
- [Llama 2: Open Foundation and Fine-Tuned Chat Models](https://mlsanity.com/paper/2307.09288/) (2023)
- [Diffusion Models Beat GANs on Image Synthesis](https://mlsanity.com/paper/2105.05233/) (2021)
- [Visual Instruction Tuning](https://mlsanity.com/paper/2304.08485/) (2023)
- [BLIP-2: Bootstrapping Language-Image Pre-training with Frozen Image Encoders and Large Language Models](https://mlsanity.com/paper/2301.12597/) (2023)
- [BLIP: Bootstrapping Language-Image Pre-training for Unified Vision-Language Understanding and Generation](https://mlsanity.com/paper/2201.12086/) (2022)
- [Qwen2-VL: Enhancing Vision-Language Model's Perception of the World at Any Resolution](https://mlsanity.com/paper/2409.12191/) (2024)
- [Generative Modeling by Estimating Gradients of the Data Distribution](https://mlsanity.com/paper/1907.05600/) (2019)
- [Scaling Up Visual and Vision-Language Representation Learning With Noisy Text Supervision](https://mlsanity.com/paper/2102.05918/) (2021)
- [InstructBLIP: Towards General-purpose Vision-Language Models with Instruction Tuning](https://mlsanity.com/paper/2305.06500/) (2023)
- [MiniGPT-4: Enhancing Vision-Language Understanding with Advanced Large Language Models](https://mlsanity.com/paper/2304.10592/) (2023)
- [Making the V in VQA Matter: Elevating the Role of Image Understanding in Visual Question Answering](https://mlsanity.com/paper/1612.00837/) (2016)
- [Qwen-VL: A Versatile Vision-Language Model for Understanding, Localization, Text Reading, and Beyond](https://mlsanity.com/paper/2308.12966/) (2023)
- [Qwen2 Technical Report](https://mlsanity.com/paper/2407.10671/) (2024)
- [Microsoft COCO Captions: Data Collection and Evaluation Server](https://mlsanity.com/paper/1504.00325/) (2015)
- [Targeted Backdoor Attacks on Deep Learning Systems Using Data Poisoning](https://mlsanity.com/paper/1712.05526/) (2017)
- [Show and Tell: A Neural Image Caption Generator](https://mlsanity.com/paper/1411.4555/) (2014)
- [Universal adversarial perturbations](https://mlsanity.com/paper/1610.08401/) (2016)
- [MiniGPT-v2: large language model as a unified interface for vision-language multi-task learning](https://mlsanity.com/paper/2310.09478/) (2023)
- [STRIP: A Defence Against Trojan Attacks on Deep Neural Networks](https://mlsanity.com/paper/1902.06531/) (2019)
- [Hidden Trigger Backdoor Attacks](https://mlsanity.com/paper/1910.00033/) (2019)
- [A Comprehensive Survey of Deep Learning for Image Captioning](https://mlsanity.com/paper/1810.04020/) (2018)
- [Clean-Label Backdoor Attacks on Video Recognition Models](https://mlsanity.com/paper/2003.03030/) (2020)
- [Narcissus: A Practical Clean-Label Backdoor Attack with Limited Information](https://mlsanity.com/paper/2204.05255/) (2022)
- [Sleeper Agent: Scalable Hidden Trigger Backdoors for Neural Networks Trained from Scratch](https://mlsanity.com/paper/2106.08970/) (2021)
- [VL-Trojan: Multimodal Instruction Backdoor Attacks against Autoregressive Visual Language Models](https://mlsanity.com/paper/2402.13851/) (2024)
- [A Survey of Safety on Large Vision-Language Models: Attacks, Defenses and Evaluations](https://mlsanity.com/paper/2502.14881/) (2025)
- [Shadowcast: Stealthy Data Poisoning Attacks Against Vision-Language Models](https://mlsanity.com/paper/2402.06659/) (2024)

## Built on by (most cited first, 3 of 3)

- [Stop When Further Reasoning Won't Help: Attention-State Adaptive Generation in Reasoning Models](https://mlsanity.com/paper/2606.15070/) (2026)
- [DIVER: Diving Deeper into Distilled Data via Expressive Semantic Recovery](https://mlsanity.com/paper/2605.12649/) (2026)
- [TraceGuard: Adaptive Multimodal Poison Filtering through Cross-Feature Rank Agreement](https://mlsanity.com/paper/2609.29099/) (2026)

## Abstract (arXiv)

Vision-Language Models (VLMs) have achieved remarkable success in tasks such as image captioning and visual question answering (VQA). However, as their applications become increasingly widespread, recent studies have revealed that VLMs are vulnerable to backdoor attacks. Existing backdoor attacks on VLMs primarily rely on data poisoning by adding visual triggers and modifying text labels, where the induced image-text mismatch makes poisoned samples easy to detect. To address this limitation, we propose the Clean-Label Backdoor Attack on VLMs via Diffusion Models (CBV), which leverages diffusion models to generate natural poisoned examples via score matching. Specifically, CBV modifies the score during the reverse generation process of the diffusion model to guide the generation of poisoned samples that contain triggered image features. To further enhance the effectiveness of the attack, we incorporate the textual information of the triggered images as multimodal guidance during generation. Moreover, to enhance stealthiness, we introduce a GradCAM-guided Mask (GM) that restricts modifications to only the most semantically important regions, rather than the entire image. We evaluate our method on MSCOCO and VQA v2 with four representative VLMs, achieving over 80% ASR while preserving normal functionality.