SafeDream: Safety World Model for Proactive Early Jailbreak Detection

• URL: http://arxiv.org/abs/2604.16824v1
• Date: Sat, 18 Apr 2026 04:31:33 GMT
• Title: SafeDream: Safety World Model for Proactive Early Jailbreak Detection
• Abstract summary: Multi-turn jailbreak attacks progressively erode LLM safety alignment across seemingly innocuous conversation turns.<n>We propose SAFEDREAM, a lightweight world-model-based framework that operates as an external module without modifying the LLM's weights.<n>On three multi-turn jailbreak benchmarks, SAFEDREAM achieves the best detection across all benchmarks (1.06-1.20 turns before compliance) while maintaining competitive false positive rates and outperforming baselines in detection quality.
• Score: 23.14126936942301
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Multi-turn jailbreak attacks progressively erode LLM safety alignment across seemingly innocuous conversation turns, achieving success rates exceeding 90% against state-of-the-art models. Existing alignment-based and guardrail methods suffer from three key limitations: they require costly weight modification, evaluate each turn independently without modeling cumulative safety erosion, and detect attacks only after harmful content has been generated. To address these limitations, we first formulate the proactive early jailbreak detection problem with a new metric, detection lead, that measures how early an attack can be detected before the LLM complies. We then propose SAFEDREAM, a lightweight world-model-based framework that operates as an external module without modifying the LLM's weights. SAFEDREAM introduces three components: (1) a safety state world model that encodes LLM hidden states into a compact safety representation and predicts how it evolves across turns, (2) CUSUM detection that accumulates weak per-turn risk signals into reliable evidence, and (3) contrastive imagination that simultaneously rolls out attack and benign futures in latent space to issue early alarms before jailbreaks occur. On three multi-turn jailbreak benchmarks (XGuard-Train, SafeDialBench, SafeMTData) against 8 baselines, SAFEDREAM achieves the best detection timeliness across all benchmarks (1.06-1.20 turns before compliance) while maintaining competitive false positive rates and outperforming baselines in detection quality.

Related papers

• SEMA: Simple yet Effective Learning for Multi-Turn Jailbreak Attacks [53.97948802255959]
  We propose a framework that trains a multi-turn attacker without relying on any existing strategies or external data.<n>Prefilling self-tuning enables usable rollouts by fine-tuning on non-refusal, well-structured, multi-turn adversarial prompts.<n>We anchor harmful intent in multi-turn jailbreaks via an intent-drift-aware reward that combines intent alignment, compliance risk, and level of detail.
  arXiv  Detail & Related papers  (2026-02-06T16:44:57Z)
• ALERT: Zero-shot LLM Jailbreak Detection via Internal Discrepancy Amplification [47.135407245022115]
  Existing detection methods mainly detect jailbreak status relying on jailbreak templates present in the training data.<n>We propose a layer-wise, module-wise, and token-wise amplification framework that progressively magnifies internal feature discrepancies between benign and jailbreak prompts.<n>Building upon these insights, we introduce ALERT, an efficient and effective zero-shot jailbreak detector.
  arXiv  Detail & Related papers  (2026-01-07T05:30:53Z)
• Jailbreaking Attacks vs. Content Safety Filters: How Far Are We in the LLM Safety Arms Race? [35.085164240720815]
  We present the first systematic evaluation of jailbreak attacks targeting large language models (LLMs)<n>Our findings suggest that prior assessments may have overestimated the practical success of these attacks.<n>There remains room to better balance recall and precision to further optimize protection and user experience.
  arXiv  Detail & Related papers  (2025-12-30T07:36:19Z)
• OmniSafeBench-MM: A Unified Benchmark and Toolbox for Multimodal Jailbreak Attack-Defense Evaluation [94.61617176929384]
  OmniSafeBench-MM is a comprehensive toolbox for multi-modal jailbreak attack-defense evaluation.<n>It integrates 13 representative attack methods, 15 defense strategies, and a diverse dataset spanning 9 major risk domains and 50 fine-grained categories.<n>By unifying data, methodology, and evaluation into an open-source, reproducible platform, OmniSafeBench-MM provides a standardized foundation for future research.
  arXiv  Detail & Related papers  (2025-12-06T22:56:29Z)
• Immunity memory-based jailbreak detection: multi-agent adaptive guard for large language models [12.772312329709868]
  Large language models (LLMs) have become foundational in AI systems, yet they remain vulnerable to adversarial jailbreak attacks.<n>We propose the Multi-Agent Adaptive Guard (MAAG) framework for jailbreak detection.<n>MAAG first extracts activation values from input prompts and compares them to historical activations stored in a memory bank for quick preliminary detection.
  arXiv  Detail & Related papers  (2025-12-03T01:40:40Z)
• Death by a Thousand Prompts: Open Model Vulnerability Analysis [0.06213771671016099]
  Open-weight models provide researchers and developers with foundations for diverse downstream applications.<n>We tested the safety and security postures of eight open-weight large language models (LLMs) to identify vulnerabilities that may impact subsequent fine-tuning and deployment.<n>Our findings reveal pervasive vulnerabilities across all tested models, with multi-turn attacks achieving success rates between 25.86% and 92.78%.
  arXiv  Detail & Related papers  (2025-11-05T07:22:24Z)
• DiffuGuard: How Intrinsic Safety is Lost and Found in Diffusion Large Language Models [50.21378052667732]
  We conduct an in-depth analysis of dLLM vulnerabilities to jailbreak attacks across two distinct dimensions: intra-step and inter-step dynamics.<n>We propose DiffuGuard, a training-free defense framework that addresses vulnerabilities through a dual-stage approach.
  arXiv  Detail & Related papers  (2025-09-29T05:17:10Z)
• Learning to Detect Unknown Jailbreak Attacks in Large Vision-Language Models [22.796169894587475]
  We propose a general framework that accurately detects unknown jailbreak attacks by shifting the focus from attack-specific learning to task-specific learning.<n>Experiments show that our method consistently higher detection AUROC on diverse unknown attacks while improving efficiency.
  arXiv  Detail & Related papers  (2025-08-08T16:13:28Z)
• ARMOR: Aligning Secure and Safe Large Language Models via Meticulous Reasoning [64.32925552574115]
  ARMOR is a large language model that analyzes jailbreak strategies and extracts the core intent.<n> ARMOR achieves state-of-the-art safety performance, with an average harmful rate of 0.002 and an attack success rate of 0.06 against advanced optimization-based jailbreaks.
  arXiv  Detail & Related papers  (2025-07-14T09:05:54Z)
• JailDAM: Jailbreak Detection with Adaptive Memory for Vision-Language Model [25.204224437843365]
  Multimodal large language models (MLLMs) excel in vision-language tasks but pose significant risks of generating harmful content.<n>Jailbreak attacks refer to intentional manipulations that bypass safety mechanisms in models, leading to the generation of inappropriate or unsafe content.<n>We introduce a test-time adaptive framework called JAILDAM to address these issues.
  arXiv  Detail & Related papers  (2025-04-03T05:00:28Z)
• Immune: Improving Safety Against Jailbreaks in Multi-modal LLMs via Inference-Time Alignment [97.38766396447369]
  Despite training-time safety alignment, Multimodal Large Language Models (MLLMs) remain vulnerable to jailbreak attacks.<n>We propose Immune, an inference-time defense framework that leverages a safe reward model through controlled decoding to defend against jailbreak attacks.
  arXiv  Detail & Related papers  (2024-11-27T19:00:10Z)
• Refuse Whenever You Feel Unsafe: Improving Safety in LLMs via Decoupled Refusal Training [67.30423823744506]
  We introduce a novel approach, Decoupled Refusal Training (DeRTa), designed to empower LLMs to refuse compliance to harmful prompts at any response position.<n>DeRTa incorporates two novel components: (1) Maximum Likelihood Estimation with Harmful Response Prefix, which trains models to recognize and avoid unsafe content by appending a segment of harmful response to the beginning of a safe response, and (2) Reinforced Transition Optimization (RTO), which equips models with the ability to transition from potential harm to safety refusal consistently throughout the harmful response sequence.
  arXiv  Detail & Related papers  (2024-07-12T09:36:33Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.