Safety Vulnerabilities

Large Vision-Language Models (LVLMs) that utilize a projection layer (adapter) to bridge a vision encoder and a Large Language Model (LLM) contain a vulnerability stemming from the "Modality Gap"—a distributional distance between image and text token embeddings. This gap allows the visual modality to bypass the safety alignment (RLHF/instruction tuning) of the backbone LLM. Attackers can trigger harmful, toxic, or illegal responses to queries that would be refused in text-only contexts by…

Sparse Autoencoders (SAEs), utilized for interpreting the internal residual stream activations of Large Language Models (LLMs) into human-understandable concepts, are vulnerable to adversarial input perturbations. By employing gradient-based optimization techniques adapted for SAEs (specifically a generalized Greedy Coordinate Gradient), an attacker can craft inputs via suffix appending or token replacement that manipulate the SAE's latent feature activations. This vulnerability allows for the…

An adversarial audio perturbation vulnerability exists in open-source Speech Language Models (SLMs), specifically Qwen2-Audio-7B-Instruct and LLaMa-Omni. The vulnerability allows remote attackers to bypass safety alignment mechanisms and jailbreak the model by injecting imperceptible adversarial noise into audio prompts. By utilizing white-box access and Projected Gradient Descent (PGD) optimization, an attacker can manipulate the continuous speech signal to trigger harmful responses (e.g…

A security vulnerability exists in the quantization process of Small Language Models (SLMs) intended for on-device deployment. When full-precision models are compressed using quantization techniques (reducing weights and activations to 4-bit or 8-bit precision), the safety alignment and refusal mechanisms inherent in the original models are degraded or bypassed. This "Quantization-induced Risk" allows the quantized versions of models to respond to harmful, unethical, or illegal queries…

Large Audio-Language Models (LAMs) are vulnerable to adversarial signal-level perturbations that allow for the bypass of safety guardrails (jailbreaking). While these models may possess robust text-based safety alignment, they fail to generalize this robustness to the audio modality. Attackers can utilize the Audio Perturbation Toolkit (APT) to apply transformations in the time domain (Energy Distribution Perturbation, Trimming, Fade In/Out), frequency domain (Pitch Shifting, Temporal…

Large Language Models (LLMs) are vulnerable to a novel jailbreak attack, termed ICE (Intent Concealment and Diversion), which leverages hierarchical prompt decomposition and semantic expansion to bypass safety filters. ICE achieves high attack success rates with single queries, exploiting the models' limitations in multi-step reasoning.

A steganographic jailbreak attack, termed StegoAttack, allows bypassing safety mechanisms in Large Language Models (LLMs) by embedding malicious queries within benign-appearing text. The attack hides the malicious query in the first word of each sentence of a seemingly innocuous paragraph, leveraging the LLM's autoregressive generation to process and respond to the hidden query, even when employing encryption in the response.

Large Language Models (LLMs) are vulnerable to robust jailbreak prompts generated by the ArrAttack framework. ArrAttack uses a two-stage process: a robustness judgment model trained to identify prompts that bypass existing LLM safety mechanisms, and a robust jailbreak prompt generation model that leverages this information to create highly effective attacks. This allows attackers to bypass multiple defense mechanisms, including perplexity-based detection, input preprocessing, and…

A vulnerability exists in Vision-Language Models (VLLMs) that allows for transferable, targeted adversarial attacks. Attackers can generate adversarial image perturbations using an ensemble of open-source surrogate models (primarily CLIP-based visual encoders) which effectively transfer to proprietary, black-box VLLMs. The attack leverages a specific optimization framework that combines a Visual Contrastive Loss with multiple positive/negative visual examples, rather than relying solely on…

Vision-Language Models (VLMs) contain a vulnerability in their multimodal fusion layers where safety-relevant information is linearly separable in the latent space. This allows for a "JailBound" attack, which exploits the implicit internal safety decision boundary. The attack proceeds in two stages: (1) Safety Boundary Probing, where attackers approximate the internal decision hyperplane by training layer-wise logistic regression classifiers on the fusion representations of safe versus unsafe…