Principled Steering via Null-space Projection for Jailbreak Defense in Vision-Language Models

• URL: http://arxiv.org/abs/2603.22094v2
• Date: Wed, 25 Mar 2026 16:02:02 GMT
• Title: Principled Steering via Null-space Projection for Jailbreak Defense in Vision-Language Models
• Abstract summary: Vision-language models can be easily induced by visual jailbreak attacks to generate harmful content.<n>We propose NullSteer, a null-space projected activation defense framework.<n>We show that NullSteer significantly reduces harmful outputs under various jailbreak attacks.
• Score: 71.11171576590134
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: As vision-language models (VLMs) are increasingly deployed in open-world scenarios, they can be easily induced by visual jailbreak attacks to generate harmful content, posing serious risks to model safety and trustworthy usage. Recent activation steering methods inject directional vectors into model activations during inference to induce refusal behaviors and have demonstrated effectiveness. However, a steering vector may both enhance refusal ability and cause over-refusal, thereby degrading model performance on benign inputs. Moreover, due to the lack of theoretical interpretability, these methods still suffer from limited robustness and effectiveness. To better balance safety and utility, we propose NullSteer, a null-space projected activation defense framework. Our method constructs refusal directions within model activations through a linear transformation: it maintains zero perturbation within the benign subspace while dynamically inducing refusal along potentially harmful directions, thereby theoretically achieving safety enhancement without impairing the model's general capabilities. Extensive experiments show that NullSteer significantly reduces harmful outputs under various jailbreak attacks (average ASR reduction over 15 percent on MiniGPT-4) while maintaining comparable performance to the original model on general benchmarks.

Related papers

• EVA: Editing for Versatile Alignment against Jailbreaks [85.18411558566851]
  EVA (Editing for Versatile Alignment against Jailbreaks) is a novel framework that pioneers the application of direct model editing for safety alignment.<n> EVA identifies and surgically edits specific neurons responsible for the model's susceptibility to harmful instructions.<n>It effectively neutralizes harmful behaviors without compromising the model's general reasoning capabilities.
  arXiv  Detail & Related papers  (2026-05-14T12:16:10Z)
• Steering Externalities: Benign Activation Steering Unintentionally Increases Jailbreak Risk for Large Language Models [62.16655896700062]
  Activation steering is a technique to enhance the utility of Large Language Models (LLMs)<n>We show that it unintentionally introduces critical and under-explored safety risks.<n>Experiments reveal that these interventions act as a force multiplier, creating new vulnerabilities to jailbreaks and increasing attack success rates to over 80% on standard benchmarks.
  arXiv  Detail & Related papers  (2026-02-03T12:32:35Z)
• Safe and Efficient In-Context Learning via Risk Control [34.917821132391374]
  Large language models (LLMs) learn new tasks from a few in-context examples.<n>LLMs can be influenced by incorrect or malicious demonstrations.<n>We propose a novel approach to limit the degree to which harmful demonstrations can degrade model performance.
  arXiv  Detail & Related papers  (2025-10-02T18:36:10Z)
• ASGuard: Activation-Scaling Guard to Mitigate Targeted Jailbreaking Attack [22.48980625853356]
  Large language models (LLMs) exhibit brittle refusal behaviors that can be circumvented by simple linguistic changes.<n>In this work, we introduce Activation-Scaling Guard (ASGuard), an insightful, mechanistically-informed framework that surgically mitigates this specific vulnerability.
  arXiv  Detail & Related papers  (2025-09-30T06:33:52Z)
• SafeSteer: Adaptive Subspace Steering for Efficient Jailbreak Defense in Vision-Language Models [25.027627636905475]
  We propose SafeSteer, a lightweight, inference-time steering framework.<n>We show that SafeSteer reduces the attack success rate by over 60% and improves accuracy on normal tasks by 1-2%.
  arXiv  Detail & Related papers  (2025-09-24T12:46:41Z)
• AlphaSteer: Learning Refusal Steering with Principled Null-Space Constraint [49.641959856967276]
  We present a theoretically grounded and empirically effective activation steering method called AlphaSteer.<n>For utility preservation, it learns to construct a nearly zero vector for steering benign data, with the null-space constraints.<n>Experiments across multiple jailbreak attacks and utility benchmarks demonstrate the effectiveness of AlphaSteer.
  arXiv  Detail & Related papers  (2025-06-08T07:03:28Z)
• AdaSteer: Your Aligned LLM is Inherently an Adaptive Jailbreak Defender [99.3105257001476]
  We propose AdaSteer, an adaptive activation steering method that adjusts model behavior based on input characteristics.<n>AdaSteer steers input representations along both the Rejection Direction (RD) and Harmfulness Direction (HD)<n>Our results highlight the potential of interpretable model internals for real-time, flexible safety enforcement in LLMs.
  arXiv  Detail & Related papers  (2025-04-13T07:39:17Z)
• Representation Bending for Large Language Model Safety [27.842146980762934]
  Large Language Models (LLMs) have emerged as powerful tools, but their inherent safety risks pose significant challenges.<n>This paper introduces RepBend, a novel approach that fundamentally disrupts the representations underlying harmful behaviors in LLMs.<n>RepBend achieves state-of-the-art performance, outperforming prior methods such as Circuit Breaker, RMU, and NPO, with up to 95% reduction in attack success rates.
  arXiv  Detail & Related papers  (2025-04-02T09:47:01Z)
• Model Tampering Attacks Enable More Rigorous Evaluations of LLM Capabilities [49.09703018511403]
  Evaluations of large language model (LLM) risks and capabilities are increasingly being incorporated into AI risk management and governance frameworks.<n>Currently, most risk evaluations are conducted by designing inputs that elicit harmful behaviors from the system.<n>We propose evaluating LLMs with model tampering attacks which allow for modifications to latent activations or weights.
  arXiv  Detail & Related papers  (2025-02-03T18:59:16Z)
• BEEAR: Embedding-based Adversarial Removal of Safety Backdoors in Instruction-tuned Language Models [57.5404308854535]
  Safety backdoor attacks in large language models (LLMs) enable the stealthy triggering of unsafe behaviors while evading detection during normal interactions. We present BEEAR, a mitigation approach leveraging the insight that backdoor triggers induce relatively uniform drifts in the model's embedding space. Our bi-level optimization method identifies universal embedding perturbations that elicit unwanted behaviors and adjusts the model parameters to reinforce safe behaviors against these perturbations.
  arXiv  Detail & Related papers  (2024-06-24T19:29:47Z)

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.