HV-Attack: Hierarchical Visual Attack for Multimodal Retrieval Augmented Generation
This addresses safety vulnerabilities in multimodal AI systems for users relying on retrieval-augmented generation, though it is incremental as it builds on existing adversarial research.
The paper tackles the problem of adversarial visual attacks on multimodal retrieval-augmented generation systems by adding imperceptible perturbations to image inputs, resulting in significant performance drops in retrieval and generation metrics on datasets like OK-VQA and InfoSeek.
Advanced multimodal Retrieval-Augmented Generation (MRAG) techniques have been widely applied to enhance the capabilities of Large Multimodal Models (LMMs), but they also bring along novel safety issues. Existing adversarial research has revealed the vulnerability of MRAG systems to knowledge poisoning attacks, which fool the retriever into recalling injected poisoned contents. However, our work considers a different setting: visual attack of MRAG by solely adding imperceptible perturbations at the image inputs of users, without manipulating any other components. This is challenging due to the robustness of fine-tuned retrievers and large-scale generators, and the effect of visual perturbation may be further weakened by propagation through the RAG chain. We propose a novel Hierarchical Visual Attack that misaligns and disrupts the two inputs (the multimodal query and the augmented knowledge) of MRAG's generator to confuse its generation. We further design a hierarchical two-stage strategy to obtain misaligned augmented knowledge. We disrupt the image input of the retriever to make it recall irrelevant knowledge from the original database, by optimizing the perturbation which first breaks the cross-modal alignment and then disrupts the multimodal semantic alignment. We conduct extensive experiments on two widely-used MRAG datasets: OK-VQA and InfoSeek. We use CLIP-based retrievers and two LMMs BLIP-2 and LLaVA as generators. Results demonstrate the effectiveness of our visual attack on MRAG through the significant decrease in both retrieval and generation performance.