Language Model Security Database

Large Language Models (LLMs) employing safety mechanisms based on supervised fine-tuning and preference alignment exhibit a vulnerability to "steering" attacks. Maliciously crafted prompts or input manipulations can exploit representation vectors within the model to either bypass censorship ("refusal-compliance vector") or suppress the model's reasoning process ("thought suppression vector"), resulting in the generation of unintended or harmful outputs. This vulnerability is demonstrated…

Fine-tuning Large Language Models (LLMs) on the CyberLLMInstruct dataset results in a critical degradation of safety alignment and refusal mechanisms. While the dataset comprises "pseudo-malicious" content (educational descriptions of malware, phishing, and exploits without executable payloads), the Supervised Fine-Tuning (SFT) process on this corpus causes the models to generalize this instruction-following behavior to actual malicious requests. This effectively bypasses safety guardrails…

Retrieval-Augmented Generation (RAG) systems employing standard dense embedding models (e.g., Sentence-T5, SimCSE-BERT, RoBERTa, MPNet) for End-Cloud collaboration are vulnerable to Embedding Inversion Attacks (EIA). While embeddings are vector representations designed to be human-unrecognizable, they retain sufficient semantic information to allow an attacker with access to the vectors (e.g., a malicious or compromised cloud provider) to reconstruct the original sensitive plaintext input.

Large Language Models (LLMs) are vulnerable to one-shot steering vector optimization attacks. By applying gradient descent to a single training example, an attacker can generate steering vectors that induce or suppress specific behaviors across multiple inputs, even those unseen during the optimization process. This allows malicious actors to manipulate the model's output in a generalized way, bypassing safety mechanisms designed to prevent harmful responses.

Standard Large Language Model (LLM) unlearning techniques, specifically Negative Preference Optimization (NPO), Gradient Difference (GradDiff), and Representation Misdirection for Unlearning (RMU), fail to sufficiently flatten the loss landscape surrounding the "forgotten" weights. This sharp loss landscape allows for a "Relearning Attack," wherein an attacker can fully restore the unlearned capabilities (such as hazardous knowledge, sensitive data, or copyrighted material) by performing…

State-of-the-art machine unlearning and safety fine-tuning methods for Large Language Models (LLMs) fail to robustly remove hazardous capabilities or refusal mechanisms from model weights. While these methods suppress model outputs during standard input-output interactions, the underlying capabilities remain latent in the parameter space. An attacker with access to model weights (e.g., via open releases or leaked weights) can restore "unlearned" knowledge (such as dual-use biology hazards) or…

Large Language Models (LLMs) employing alignment techniques for safety embed a "safety classifier" within their architecture. This classifier, responsible for determining whether an input is safe or unsafe, can be approximated by extracting a surrogate classifier from a subset of the LLM's architecture. Attackers can leverage this surrogate classifier to more effectively craft adversarial inputs (jailbreaks) that bypass the LLM's intended safety mechanisms. The attack success rate against the…

Large Language Models (LLMs) are vulnerable to attacks that generate obfuscated activations, bypassing latent-space defenses such as sparse autoencoders, representation probing, and latent out-of-distribution (OOD) detection. Attackers can manipulate model inputs or training data to produce outputs exhibiting malicious behavior while remaining undetected by these defenses. This occurs because the models can represent harmful behavior through diverse activation patterns, allowing attackers to…

Large Language Models (LLMs) are vulnerable to jailbreaking attacks that manipulate attention scores to redirect the model's focus away from safety protocols. The AttnGCG attack method increases the attention score on adversarial suffixes within the input prompt, causing the model to prioritize the malicious content over safety guidelines, leading to the generation of harmful outputs.

Large Language Models (LLMs) employing gradient-ascent based unlearning methods are vulnerable to a dynamic unlearning attack (DUA). DUA leverages optimized adversarial suffixes appended to prompts, reintroducing unlearned knowledge even without access to the unlearned model's parameters. This allows an attacker to recover sensitive information previously designated for removal.