Skip to content
#small language model Book Open access

DIRE: Dual-grained Internal Refusal Estimation for VLM Jailbreak Detection

Nov 2026 · Proceedings of the 34th ACM International Conference on Multimedia · 0 citations · 12 references

TL;DR

Dual-grained Internal Refusal Estimation (DIRE), a training-free detection framework that jointly probes resilient Attention routing and FFN semantic memory within effective semantic layers to detect multimodal jailbreaks in a training-free manner, is introduced.

Abstract

While Vision-Language Models (VLMs) excel at multimodal tasks, they are increasingly vulnerable to sophisticated multimodal jailbreak attacks. Conventional defenses typically introduce significant training overhead, risk degrading model capabilities, or cause severe inference latency via external modules. Additionally, current internal state detection methods rely on coarse-grained feature analysis, often failing to capture the micro-level conflicts generated by jailbreak attacks. To overcome these challenges, we analyze multimodal jailbreaks through mechanistic interpretability. We demonstrate that successful attacks typically do not eradicate intrinsic safety knowledge but instead cause a widespread attenuation of refusal signals. Even when global safety alignment is bypassed, a small, robust subset of Attention and FFN components continues to exhibit directional resistance against malicious queries. Building upon this empirical finding, we introduce Dual-grained Internal Refusal Estimation (DIRE), a training-free detection framework. Unlike prior internal monitors that are either macro-only or attention-only, DIRE jointly probes resilient Attention routing and FFN semantic memory within effective semantic layers to detect multimodal jailbreaks in a training-free manner. Comprehensive evaluations on mainstream VLMs and diverse benchmarks demonstrate that DIRE achieves competitive and strong performance, with an average AUROC of 0.9800. For better reproducibility, the anonymized code is available at: https://github.com/xiyuzzz-njust/DIRE.

Read PDF

Similar papers

#artificial intelligence Preprint Oct 2026

BRANCH: Bypassing Multi-Scanner AI Guardrails

This work proposes BRANCH, a bypassing methodology designed for multi-scanner guardrail systems that achieves 100% attack success rate across 6 guardrail systems in 120 scenarios with 72% fewer queries and 4.5x reduced wallclock time compared to established techniques, while preserving semantic meaning within the bypas...

William Hackett, Peter Garraghan · 0 citations
#natural language process... Preprint Nov 2026

Beyond AUROC: Deployment-Aware Evaluation of Model-Internal Jailbreak Detectors

Large language models (LLMs) are increasingly deployed in user-facing systems, where jailbreak attacks can bypass safety controls and induce harmful and policy-violating outputs. Recent model-internal detectors use hidden representations, gradients, or internal signals to identify such prompts, but their practical secu...

Mohammad Tehabsim, Basil Al-Housani, Nour Alhussien et al. · 0 citations
Preprint Nov 2026

JailBound: FOL-Guided Jailbreak Evaluation for Exploring LLM Behavioral Boundaries

JailBound is presented, a unified jailbreak evaluation framework that combines automated benchmark construction with first-order loss (FOL)-guided dual-branch search and organizes threats by risk category, application domain, and attack type.

Fa-Zong Wu, Xin Wang, Ming Yang et al. · 0 citations

Decoding Guardrails: XAI-Guided Perturbation Analysis of Prompt Injection Detection

An exploratory case study that applies explainable artificial intelligence techniques to analyze how Prompt Guard 2 distinguishes malicious from benign prompts finds that Prompt Guard 2's decisions rely on the cumulative contribution of many tokens rather than a few dominant ones, yet saliency-guided synonym substituti...

Fernando Outeda, Gustavo Betarte, J. Campo et al. · 0 citations
#artificial intelligence Preprint Oct 2026

Logic Hijacking of RAG: Single-Document Poisoning via Local Optimization

Single-document corpus poisoning under black-box access poses a severe threat to Retrieval-Augmented Generation (RAG). However, existing attacks predominantly rely on explicit factual refutations or imperative instruction overrides. Modern aligned LLMs often arbitrate such conflicts conservatively, rejecting overt cont...

Hang-Yi Zhao, Bo-Nan Zhang, Fu-Sen Guo et al. · 0 citations
Preprint Oct 2026

Combining Long-Term Memory and LLMs Comes at a High Cost: From Prompt Injection to Persistent Compromise

The widespread adoption of generative artificial intelligence in recent years has brought a parallel rise in prompt injection and jailbreak attacks. By embedding malicious instructions, these exploits manipulate language models to bypass intended behavior and operational boundaries. In traditional chatbots, such attack...

Pedro H. Barcha Correia, Matheus Azevedo de Sá, P. D. de Senna et al. · 0 citations

Related blog posts

Microsoft Research Blog Nov 16, 2026

TokenEconomizer: Learning from Experience to Make AI Coding Agents More Efficient

AI coding agents are increasingly help developers fix bugs and build features largely on their own, but every interaction with their models consumes tokens. In each session, an agent sends instructions, tool descriptions, and session history to its model. Most efforts to reduce this cost focus on individual tasks. Yet developers return to the same […] The post TokenEconomizer: Learning from Experience to Make AI Coding Agents More Efficient appeared first on Microsoft Research.

We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.