PiCo: Jailbreaking Multimodal Large Language Models via Pictorial Code Contextualization

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Liu, Aofan, Tang, Lulu, Pan, Ting, Yin, Yuguo, Wang, Bin, Yang, Ao
Formato: Preprint
Publicado: 2025
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866918156876906496
author Liu, Aofan
Tang, Lulu
Pan, Ting
Yin, Yuguo
Wang, Bin
Yang, Ao
author_facet Liu, Aofan
Tang, Lulu
Pan, Ting
Yin, Yuguo
Wang, Bin
Yang, Ao
contents Multimodal Large Language Models (MLLMs), which integrate vision and other modalities into Large Language Models (LLMs), significantly enhance AI capabilities but also introduce new security vulnerabilities. By exploiting the vulnerabilities of the visual modality and the long-tail distribution characteristic of code training data, we present PiCo, a novel jailbreaking framework designed to progressively bypass multi-tiered defense mechanisms in advanced MLLMs. PiCo employs a tier-by-tier jailbreak strategy, using token-level typographic attacks to evade input filtering and embedding harmful intent within programming context instructions to bypass runtime monitoring. To comprehensively assess the impact of attacks, a new evaluation metric is further proposed to assess both the toxicity and helpfulness of model outputs post-attack. By embedding harmful intent within code-style visual instructions, PiCo achieves an average Attack Success Rate (ASR) of 84.13% on Gemini-Pro Vision and 52.66% on GPT-4, surpassing previous methods. Experimental results highlight the critical gaps in current defenses, underscoring the need for more robust strategies to secure advanced MLLMs.
format Preprint
id arxiv_https___arxiv_org_abs_2504_01444
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle PiCo: Jailbreaking Multimodal Large Language Models via Pictorial Code Contextualization
Liu, Aofan
Tang, Lulu
Pan, Ting
Yin, Yuguo
Wang, Bin
Yang, Ao
Cryptography and Security
Artificial Intelligence
Multimodal Large Language Models (MLLMs), which integrate vision and other modalities into Large Language Models (LLMs), significantly enhance AI capabilities but also introduce new security vulnerabilities. By exploiting the vulnerabilities of the visual modality and the long-tail distribution characteristic of code training data, we present PiCo, a novel jailbreaking framework designed to progressively bypass multi-tiered defense mechanisms in advanced MLLMs. PiCo employs a tier-by-tier jailbreak strategy, using token-level typographic attacks to evade input filtering and embedding harmful intent within programming context instructions to bypass runtime monitoring. To comprehensively assess the impact of attacks, a new evaluation metric is further proposed to assess both the toxicity and helpfulness of model outputs post-attack. By embedding harmful intent within code-style visual instructions, PiCo achieves an average Attack Success Rate (ASR) of 84.13% on Gemini-Pro Vision and 52.66% on GPT-4, surpassing previous methods. Experimental results highlight the critical gaps in current defenses, underscoring the need for more robust strategies to secure advanced MLLMs.
title PiCo: Jailbreaking Multimodal Large Language Models via Pictorial Code Contextualization
topic Cryptography and Security
Artificial Intelligence
url https://arxiv.org/abs/2504.01444