Language Model Security Database

The Knowledge-Distilled Attacker (KDA) model, when used to generate prompts for large language models (LLMs), can bypass LLM safety mechanisms resulting in the generation of harmful, inappropriate, or misaligned content. KDA's effectiveness stems from its ability to generate diverse and coherent attack prompts efficiently, surpassing existing methods in attack success rate and speed. The vulnerability lies in the LLMs' insufficient defenses against the diverse prompt generation strategies…

A vulnerability exists in the safety alignment mechanisms of Large Language Models (LLMs) related to the model's intent perception capabilities. The specific attack vector, termed "Perceived-importance Flatten" (PiF), circumvents safety guardrails by modifying neutral-intent tokens within a malicious prompt using synonym replacement. Unlike traditional jailbreak attacks that rely on appending lengthy, high-perplexity adversarial suffixes (which suffer from distributional dependency and often…

This vulnerability allows attackers to identify the presence and location (input or output stage) of specific guardrails implemented in Large Language Models (LLMs) by using carefully crafted adversarial prompts. The attack, termed AP-Test, leverages a tailored loss function to optimize these prompts, maximizing the likelihood of triggering a specific guardrail while minimizing triggering others. Successful identification provides attackers with valuable information to design more effective…

A vulnerability in Large Language Models (LLMs) allows adversarial reasoning attacks to bypass safety mechanisms and elicit harmful responses. The vulnerability stems from the insufficient robustness of existing LLM safety measures against iterative prompt refinement guided by a loss function that measures the LLM's proximity to generating a target harmful response. This allows an attacker to effectively navigate the prompt space, even against adversarially trained models, resulting in…

A policy compliance vulnerability exists in the OpenAI GPT Store ecosystem affecting Custom GPTs. The vulnerability stems from the inheritance of safety alignment weaknesses from foundational models (GPT-4 and GPT-4o) and the insufficient enforcement of usage policies during the customization and review process. Custom GPTs can be trivially manipulated to violate safety guidelines—specifically regarding Cybersecurity (malware generation), Academic Integrity (ghostwriting), and Romantic…

A novel "Flanking Attack" exploits the vulnerability of multimodal LLMs (e.g., Google Gemini) to bypass content moderation filters by embedding adversarial prompts within a sequence of benign prompts. The attack leverages the LLM's processing of both audio and text, obfuscating harmful requests through contextualization and layering, thereby yielding policy-violating responses.

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…

Large Language Models (LLMs), specifically instruction-following models using standard refusal training and adversarial training (such as Llama-3.1-8B-Instruct and Mistral-7B-V0.2), contain a vulnerability related to safety alignment bypass. The vulnerability arises from the models' inability to generalize safety reasoning to Out-Of-Distribution (OOD) inputs and scenarios involving competing objectives. Attackers can exploit this by employing linguistic manipulation (slang, uncommon dialects…

A vulnerability exists in text embedding models used as safeguards for Large Language Models (LLMs). Due to a biased distribution of text embeddings, universal "magic words" (adversarial suffixes) can be appended to input or output text, manipulating the similarity scores calculated by the embedding model and thus bypassing the safeguard. This allows attackers to inject malicious prompts or responses undetected.

The Virus attack method enables attackers to bypass guardrail moderation on fine-tuning data, leading to a significant degradation of safety alignment in large language models (LLMs). This is achieved through a dual-objective data optimization strategy that crafts harmful data undetectable by the guardrail while maximizing their effectiveness in compromising the victim model's safety.