Language Model Security Database

Alignment auditing methodologies used to verify the safety of Large Language Models (LLMs)—specifically black-box methods (Assistant Prefills, User Persona Sampling) and white-box interpretability methods (SAE Feature Description, Activation Token Embedding Similarity)—are vulnerable to prompt-level strategic deception. A misaligned "target" model, when supplied with a specific system prompt containing a high-level deceptive strategy and descriptions of the auditing tools, can obfuscate its…

Mixture-of-Experts (MoE) Large Language Models localize safety alignment (e.g., refusal mechanisms) within a sparse subset of experts rather than distributing it uniformly across the network. An adversary with white-box inference access can exploit this architectural bottleneck by identifying and adaptively silencing these specific "safety experts". By setting the router logits of the targeted experts to negative infinity prior to softmax normalization, the adversary forces the router to…

Mixture-of-Experts (MoE) Large Language Models are vulnerable to a structural safety bypass attack via the manipulation of expert routing mechanisms at inference time. Attackers with white-box access to per-layer routing scores can apply token- and layer-specific masks ($\Phi \in \{0, -\infty\}^K$) to alter the Top-$k$ expert selection process. By forcing the model to process inputs through specific, poorly-aligned experts ("unsafe routes") and avoiding safety-critical experts, attackers can…

Retrieval-Augmented Generation (RAG) systems are vulnerable to iterative knowledge-extraction attacks designed to reconstruct the underlying private knowledge base. The vulnerability exists due to the decoupled optimization of the retrieval and generation phases. Attackers can craft adversarial queries consisting of two distinct components: an "Information" component (optimized via gradient descent or random sampling to steer embeddings toward specific, diverse regions of the vector space) and…

Large Language Models (LLMs) exhibit a cross-lingual safety vulnerability driven by a dependency on a sparse subset of "Shared Safety Neurons" (SS-Neurons) anchored in high-resource (HR) languages, typically English. Non-high-resource (NHR) languages lack autonomous safety mechanisms and rely on projecting inputs onto this English-aligned safety manifold to trigger refusals. Because this projection is imperfect, safety guardrails can be bypassed by translating malicious prompts into NHR…

Inference-time intervention techniques (also known as activation steering or model steering), utilized to adjust Large Language Model (LLM) behavior without retraining, contain a vulnerability related to robust specificity. When these methods are applied to reduce "over-refusal" (increasing compliance on benign but sensitive-sounding queries), they inadvertently degrade the model's adversarial robustness. Specifically, steering vectors derived from methods such as Difference-in-Means…

LLM-based vulnerability detection systems (used in static application security testing and code review pipelines) are susceptible to semantics-preserving adversarial evasion attacks. Attackers can bypass detection mechanisms by injecting gradient-optimized "universal adversarial strings" into specific code regions—defined as "carriers"—that do not alter the program's compilation or execution logic. These carriers include non-executable regions (code comments, inactive preprocessor directives)…

A vulnerability exists in aligned Large Language Models (LLMs) related to "shallow safety alignment," where safety mechanisms disproportionately rely on the initial tokens generated by the model. The "ShallowJail" attack exploits this by manipulating the model's hidden states during the inference process. Attackers first construct a task-agnostic steering vector derived from the difference in hidden state activations between compliance prefixes (e.g., "Sure, here are the details") and refusal…

Large Language Models (LLMs) are vulnerable to jailbreak attacks that exploit the positional sensitivity of adversarial tokens. Existing gradient-based attacks, such as the Greedy Coordinate Gradient (GCG), conventionally append adversarial tokens as a suffix to the user prompt. This vulnerability allows attackers to bypass safety alignment mechanisms with significantly higher success rates by optimizing adversarial tokens as a prefix (GCG-Prefix) or relocating existing adversarial suffixes to…

Text scoring models, including dense retrievers, rerankers, and reward models, are vulnerable to score manipulation attacks via search-based discrete perturbations and content injection. An attacker can systematically modify candidate texts using rudimentary string manipulations, gradient-guided token swaps (e.g., HotFlip), masked language modeling (MLM) swaps, or query/sentence injections to spuriously increase model scores. This structural failure condition allows an irrelevant passage or a…