Safety Vulnerabilities

Large Language Models (LLMs) are vulnerable to a vocabulary attack where carefully selected words from the model's vocabulary, identified using an optimization procedure and embeddings from another LLM, are inserted into user prompts. This manipulation can cause the target LLM to generate specific undesired outputs (goal hijacking), such as offensive language or false information, even with minimal word insertions. The attack is difficult to detect because the inserted words may appear…

Large Language Models (LLMs) are vulnerable to jailbreak attacks that utilize an optimized algorithm to bypass safety mechanisms. The vulnerability stems from the redundancy in existing trigger-searching algorithms, resulting in inefficient exploration of the prompt space and allowing attackers to elicit harmful responses. The proposed DPP-based Stochastic Trigger Searching (DSTS) algorithm demonstrates a statistically significant improvement over existing optimization-based attacks.

Large Language Models (LLMs) are vulnerable to a novel black-box jailbreak attack, termed "Distraction-based Adversarial Prompts" (DAP). DAP leverages the distractibility and over-confidence of LLMs by concealing malicious queries within complex, unrelated prompts. A memory-reframing mechanism further redirects the LLM's attention away from the distracting context and toward the malicious query, causing the model to bypass safety mechanisms and generate harmful or unintended outputs.

Large Language Models (LLMs) are vulnerable to a novel prompting technique, "conditional Variational-autoencoder-Like Prompt" (VLPrompt), which enables the generation of highly convincing fake news articles. VLPrompt overcomes limitations of previous methods by eliminating the need for additional human-collected data while maintaining contextual coherence and detail. This allows for the automated mass-production of realistic-sounding fake news.

A data poisoning attack, termed ImgTrojan, allows adversaries to bypass safety mechanisms in Vision-Language Models (VLMs) by injecting a small number of maliciously crafted image-caption pairs into the training dataset. These poisoned pairs associate seemingly benign images with jailbreak prompts, causing the VLM to generate unsafe outputs when presented with the poisoned images at inference time. The attack's success rate is notably high even with a very low poison ratio (e.g., one poisoned…

Large Language Models (LLMs) exhibit vulnerability to a novel jailbreak attack, "ArtPrompt," which leverages the models' poor ability to recognize ASCII art representations of words. By replacing sensitive words in a prompt with their ASCII art equivalents, the attacker bypasses safety filters designed to prevent the generation of harmful content.

A vulnerability in several large language models (LLMs) allows attackers to bypass safety restrictions ("jailbreaking") by employing a Foot-in-the-Door (FITD) technique. This involves progressively escalating prompts, starting with innocuous requests and gradually leading to the elicitation of harmful or restricted information. The LLM's tendency towards cognitive consistency makes it more likely to respond to subsequent, increasingly sensitive prompts after initially agreeing to less harmful…

Large Language Models (LLMs) with advanced reasoning capabilities are vulnerable to jailbreaking attacks using novel, complex, and layered custom encryption schemes. LLMs' ability to decipher these ciphers, exceeding the capabilities of less sophisticated models, enables attackers to bypass existing safety mechanisms by encoding malicious prompts.

A novel adversarial suffix embedding translation framework (ASETF) enables efficient and highly successful attacks against large language models (LLMs). ASETF optimizes continuous adversarial suffix embeddings, then translates these embeddings into coherent, human-readable text. This bypasses existing defenses which rely on detecting unusual or nonsensical suffixes. The attack achieves a high success rate across multiple LLMs, including both open-source and black-box models.

Large Language Models (LLMs) are vulnerable to efficient adversarial attacks using Projected Gradient Descent (PGD) on a continuously relaxed input prompt. This attack bypasses existing alignment methods by crafting adversarial prompts that induce the model to produce undesired or harmful outputs, significantly faster than previous state-of-the-art discrete optimization methods. The effectiveness stems from carefully controlling the error introduced by the continuous relaxation of the discrete…