Customization under Fire: Plugin Poisoning in Text-To-Image Ecosystem

The prosperity of text-to-image T2I models has fostered a vibrant share-and-play ecosystem centered on Low-Rank Adaptation LoRA plugins, which allow users to customize and share model capabilities with ease. This democratization, however, comes with a hidden but severe security risk. Malicious...

Refusal Before Decoding: Detecting and Exploiting Refusal Signals in Intermediate LLM Activations

In this paper, we investigate whether refusal behavior can be predicted from LLM intermediate activations before decoding using linear probes trained on residual stream activations at each transformer block. We find that refusal is linearly decodable well before the final layer, indicating that...

CVE-2026-7020

CVE-2026-7020 affects Ollama up to version 0.20.2. The vulnerability lies in the digestToPath function (x/imagegen/transfer/transfer.go) where manipulating the digest enables path traversal. The attack can be performed remotely and is described as high complexity with a documented PoC/exploit. Co...

Ollama 路径遍历漏洞

Ollama is an open-source tool developed by Ollama that can be run locally, used for managing and customizing large language models. Ollama versions 0.20.2 and earlier had a path traversal vulnerability. This vulnerability stemmed from the operation of the digestToPath function in the Tensor Model...

PT-2026-35201

A security flaw has been discovered in Ollama up to 0.20.2. This affects the function digestToPath of the file x/imagegen/transfer/transfer.go of the component Tensor Model Transfer Handler. The manipulation of the argument digest results in path traversal. The attack may be performed from remote...

Multi-Faceted Attack: Exposing Cross-Model Vulnerabilities in Defense-Equipped Vision-Language Models

The growing misuse of Vision-Language Models VLMs has led providers to deploy multiple safeguards, including alignment tuning, system prompts, and content moderation. However, the real-world robustness of these defenses against adversarial attacks remains underexplored. We introduce Multi-Faceted...

DAVSP: Safety Alignment for Large Vision-Language Models Via Deep Aligned Visual Safety Prompt

Large Vision-Language Models LVLMs have achieved impressive progress across various applications but remain vulnerable to malicious queries that exploit the visual modality. Existing alignment approaches typically fail to resist malicious queries while preserving utility on benign ones effectivel...

R1dacted: Investigating Local Censorship in DeepSeek'S R1 Language Model

DeepSeek recently released R1, a high-performing large language model LLM optimized for reasoning tasks. Despite its efficient training pipeline, R1 achieves competitive performance, even surpassing leading reasoning models like OpenAI's o1 on several benchmarks. However, emerging reports suggest...

Revealing Weaknesses in Text Watermarking through Self-Information Rewrite Attacks

Text watermarking aims to subtly embed statistical signals into text by controlling the Large Language Model LLM's sampling process, enabling watermark detectors to verify that the output was generated by the specified model. The robustness of these watermarking algorithms has become a key factor...