Image-based Prompt Injection: Hijacking Multimodal LLMs through Visually Embedded Adversarial Instructions

Researchers have developed Image-based Prompt Injection (IPI), a black-box attack that embeds adversarial instructions into natural images to manipulate multimodal AI models. Testing on GPT-4-turbo achieved up to 64% attack success rate, demonstrating a significant security vulnerability in vision-language AI systems.

• →Image-based Prompt Injection can reliably manipulate multimodal AI models by hiding malicious instructions in images.
• →The attack achieved up to 64% success rate against GPT-4-turbo while remaining hidden from human perception.
• →The technique uses segmentation-based region selection and background-aware rendering to conceal adversarial prompts.
• →This vulnerability affects all multimodal large language models that process both vision and text inputs.
• →The findings highlight an urgent need for new defense mechanisms against multimodal prompt injection attacks.