Pith. sign in

REVIEW 6 cited by

Shaping the Safety Boundaries: Understanding and Defending Against Jailbreaks in Large Language Models

Not yet reviewed by Pith; the record is open.

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2412.17034 v2 pith:VPOOLXTY submitted 2024-12-22 cs.CL

classification cs.CL
keywords boundarydefensesafetyfindharmfullayerslessllms
verification ladder T0 review T1 audit T2 compute T3 formal
0 comments
read the original abstract

Jailbreaking in Large Language Models (LLMs) is a major security concern as it can deceive LLMs to generate harmful text. Yet, there is still insufficient understanding of how jailbreaking works, which makes it hard to develop effective defense strategies. We aim to shed more light into this issue: we conduct a detailed large-scale analysis of seven different jailbreak methods and find that these disagreements stem from insufficient observation samples. In particular, we introduce \textit{safety boundary}, and we find that jailbreaks shift harmful activations outside that safety boundary, where LLMs are less sensitive to harmful information. We also find that the low and the middle layers are critical in such shifts, while deeper layers have less impact. Leveraging on these insights, we propose a novel defense called \textbf{Activation Boundary Defense} (ABD), which adaptively constrains the activations within the safety boundary. We further use Bayesian optimization to selectively apply the defense method to the low and the middle layers. Our experiments on several benchmarks show that ABD achieves an average DSR of over 98\% against various forms of jailbreak attacks, with less than 2\% impact on the model's general capabilities.

Discussion (0). Sign in to comment.

Forward citations

Cited by 6 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. GhostPrompt: Cross-Image Adversarial Prompt for Vision-Language Models

    cs.CR 2026-07 conditional novelty 6.0 of 10

    GhostPrompt is a universal adversarial text suffix that, after one optimization, steers VLMs to attacker-chosen outputs across diverse unseen images, reporting >30% ASR gains over prior prompt attacks.

  2. LLM in the Middle: A Systematic Review of Threats and Mitigations to Real-World LLM-based Systems

    cs.CR 2025-09 conditional novelty 6.0 of 10

    A systematic review that categorizes LLM threats, severity scores, and mitigations across development and operation life cycles and multiple deployment scenarios.

  3. Con Instruction: Universal Jailbreaking of Multimodal Large Language Models via Non-Textual Modalities

    cs.CR 2025-05 conditional novelty 6.0 of 10

    Con Instruction embeds harmful textual instructions into adversarial images or audio by aligning their representations, achieving successful jailbreaks on several vision- and audio-language models.

  4. Audio Jailbreak: An Open Comprehensive Benchmark for Jailbreaking Large Audio-Language Models

    cs.SD 2025-05 conditional novelty 6.0 of 10

    AJailBench is an open benchmark showing that large audio-language models can be jailbroken through TTS-converted text attacks and through subtle acoustic perturbations that preserve speech semantics.

  5. VSCBench: Bridging the Gap in Vision-Language Model Safety Calibration

    cs.IR 2025-05 conditional novelty 5.0 of 10

    A new benchmark, VSCBench, measures oversafety and undersafety in vision-language models and shows that most models, including proprietary ones, are miscalibrated on at least one safety dimension.

  6. Probing the Robustness of Large Language Models Safety to Latent Perturbations

    cs.LG 2025-06 conditional novelty 4.0 of 10

    Randomized noise injected into hidden layers bypasses safety refusals in 12 open LLMs, and layer-wise adversarial training on the resulting benchmark reduces the attack's success.

Pith tools