Language Model Security Database

Post-hoc Large Language Model (LLM) unlearning and guardrailing mechanisms (specifically In-Context Unlearning [ICUL] and standard prompt-based Guardrailing) are vulnerable to information leakage attacks via "Target Masking" and indirect referencing. These systems rely on superficial semantic matching to suppress "forget sets" (specific entities or concepts). Attackers can bypass these restrictions by querying associated properties, relationships, or pseudonyms rather than the explicit target…

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) used in hate speech detection systems are vulnerable to adversarial attacks and model stealing, resulting in evasion of hate speech detection. Adversarial attacks modify hate speech text to evade detection, while model stealing creates surrogate models that mimic the target system's behavior.

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 a novel agentic-based red-teaming attack, PrivAgent, which uses reinforcement learning to generate adversarial prompts. These prompts can extract sensitive information, including system prompts and portions of training data, from target LLMs even with existing guardrail defenses. The attack leverages a custom reward function based on a normalized sliding-window word edit similarity metric to guide the learning process, enabling it to overcome the…

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.

Jailbreaking vulnerabilities in Large Language Models (LLMs) used in Retrieval-Augmented Generation (RAG) systems allow escalation of attacks from entity extraction to full document extraction and enable the propagation of self-replicating malicious prompts ("worms") within interconnected RAG applications. Exploitation leverages prompt injection to force the LLM to return retrieved documents or execute malicious actions specified within the prompt.

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.

A Cross-Prompt Injection Attack (XPIA) can be amplified by appending a Greedy Coordinate Gradient (GCG) suffix to the malicious injection. This increases the likelihood that a Large Language Model (LLM) will execute the injected instruction, even in the presence of a user's primary instruction, leading to data exfiltration. The success rate of the attack depends on the LLM's complexity; medium-complexity models show increased vulnerability.