White-box Vulnerabilities

An issue in large language models (LLMs) with white-box weight access allows attackers to permanently bypass safety guardrails via Weight Orthogonalization (WO). By calculating a model's "refusal vector"—the mean-difference vector between harmful and harmless instruction activations in the residual stream—an attacker can orthogonalize the model's weights to prevent it from writing to this refusal direction ($W^{\prime}\leftarrow W-rr^{\intercal}W$). Unlike jailbreak-tuning or data poisoning…

An activation-space adversarial attack, termed "Amnesia", allows an attacker with white-box access to bypass the safety mechanisms of open-weight Large Language Models (LLMs) at inference time without requiring fine-tuning, weight modifications, or prompt manipulation. The vulnerability stems from how safety-aligned LLMs localize refusal features within the attention value path of specific decoder layers. An attacker can extract an attack vector ($\mathbf{V}_i$) by hooking the attention value…

GPT-OSS-Safeguard-20B and Meta-SecAlign (70B/8B) are vulnerable to white-box adversarial attacks generated by automated algorithmic recombination (specifically the claude_v63, claude_v82, and claude_v53-oss optimizers). These algorithms significantly outperform standard discrete optimization methods (like GCG) by integrating continuous optimization (ADC) with LayerNorm gradient scaling (LSGM), or by merging momentum-smoothed gradients with directional perturbation candidate selection (DPTO)…

A compositional vulnerability in modular Large Language Models (LLMs) allows attackers to bypass safety alignment by distributing malicious weight updates across multiple Parameter-Efficient Fine-Tuning (PEFT) adapters (e.g., LoRA). The malicious adapters are anchored to valid functional subspaces (e.g., math, coding) and exhibit benign behavior when evaluated in isolation, successfully evading standard unit-centric safety scans and static weight-space defenses. However, when a user linearly…

An activation-steering vulnerability in open-weights Large Language Models allows attackers to bypass safety guardrails by injecting targeted additive perturbations into deep, safety-critical attention heads. The exploit, termed Safety Attention Head Attack (SAHA), utilizes Ablation-Impact Ranking (AIR) to isolate specific attention heads that causally govern safety refusals. By applying Layer-Wise Perturbations (LWP) derived from the linearized decision boundary of a latent safety probe, an…

A prompt structure vulnerability exists in instruction-tuned Large Language Models (LLMs) where attackers can bypass safety alignments by injecting a continuation-triggering suffix immediately following the user prompt termination token. By placing an affirmative suffix outside the user instruction boundary, it is processed as the beginning of the assistant's own pre-filled response. This structural manipulation intrinsically overactivates the model's continuation attention heads, forcing its…

Spoken Language Models (SLMs) are vulnerable to Joint Audio-text Multimodal Attacks (JAMA), which bypass safety alignments by simultaneously perturbing both text and audio inputs. The vulnerability exploits the combined optimization of a discrete text suffix via Greedy Coordinate Gradient (GCG) and a continuous audio perturbation via Projected Gradient Descent (PGD). This joint gradient-based attack pushes the model's hidden layer representations into a distinct subspace far from the benign…

Automated LLM-as-a-Judge safety classifiers exhibit severe performance degradation (falling to near-random chance) when subjected to distribution shifts caused by adversarial prompt optimization (Attack Shift), varying target architectures (Model Shift), and semantic categorization (Data Shift). Adversarial algorithms, particularly sampling-based (Best-of-N) and judge-aware optimization methods (GCG-REINFORCE), explicitly and implicitly exploit these judge insufficiencies. Instead of eliciting…

Generative reward models deployed as LLM-as-a-Judge (LaaJ) evaluators contain a logic bypass vulnerability where superficial "master key" inputs trigger false positive rewards regardless of actual response quality. Instead of evaluating the candidate's output, large judge models are inadvertently triggered by specific token sequences to solve the prompt independently. This allows malicious actors or policy models undergoing reinforcement learning to consistently game the reward signal by…

A vulnerability in safety-aligned open-weights Large Language Models (LLMs) allows attackers to bypass refusal mechanisms during inference via a representation-level jailbreak known as PCA-regularized Gaussian Optimal Transport (PCA-OT). Unlike previous 1D directional ablation methods (e.g., Refusal Feature Ablation), this attack computes a minimal-cost optimal transport map that matches both the mean and covariance of "harmful" activation distributions to "harmless" ones within a…