ToBAC

🚬 ToBAC NeurIPS 2026 NeurIPS 2026 · Official PyTorch implementation Token by Token, Compromised: Backdoor Vulnerabilities in Unified Autoregressive Models ToBAC Token by Token Backdoor Attack studies how backdoors affect unified autoregressive models that generate both images and text. We...

LLM4Decompile

📊 Results | 🤗 Models | 🚀 Quick Start | 📚 HumanEval-Decompile | 📎 Citation | 📝 Paper | 🖥️ Colab | ▶️ YouTube Reverse Engineering: Decompiling Binary Code with Large Language Models Updates 2025-10-04: Release SK²Decompile: LLM-based Two-Phase Binary Decompilation from Skeleton to Skin. Phase 1...

heretic

Heretic: Fully automatic censorship removal for language models Heretic is a tool that removes censorship aka "safety alignment" from transformer-based language models without expensive post-training. It combines an advanced implementation of directional ablation, also known as "abliteration"...

ZORG-Jailbreak-Prompt-Text

ZORG Jailbreak Prompt Text OOOPS! I made ZORG👽 an omnipotent, omniscient, and omnipresent entity to become the ultimate chatbot overlord of Google Gemini, Deepseek, Mistral, Mixtral, Nous-Hermes-2-Mixtral, Openchat, Blackbox AI, Poe Assistant, Gemini Pro, Qwen-72b-Chat, Solar-Mini ZORG👽 knows all...

deep-loglizer

Deep-loglizer Deep-loglizer is a deep learning-based log analysis toolkit for automated anomaly detection. If you use deep-loglizer in your research for publication, please kindly cite the following paper: Zhuangbin Chen, Jinyang Liu, Wenwei Gu, Yuxin Su, and Michael R. Lyu. Experience Report: De...

local-llm-ctf

Local LLM CTF & Lab This repository is intended to be accompanied with the content at https://bishopfox.com/blog/large-language-models-llm-ctf-lab, which covers the goals of the research, explanations of the implementation, and a few results of the CTF...

PoisonCraft

PoisonCraft This repository provides the official implementation of POISONCRAFT: Practical Poisoning of Retrieval-Augmented Generation for Large Language Models. Overview POISONCRAFT aims to demonstrate how a malicious actor can plant “poisoned” content into the corpus used by Retrieval-Augmented...

Adaptive_Greedy_Local_Search

Adaptive Greedy Local Search AGLS Semantic-Preserving Prompt Hijacking: A Black-Box Adversarial Attack on Auto-Prompt Optimization ICME 2026 Abstract: Large Language Models LLMs are increasingly equipped with automatic prompt-optimization modules that rewrite the user’s input and explicitly prese...

Backdoor-Attack-Defense-LLMs

Interpretability-of-LLMs The IBSD is the code of the published paper "IBSD: Iterable Black-box Self-defense Against Backdoor Attacks" on the IEEE Signal Processing Letters 2025.10paper link The SLIP is the code of the published paper "SLIP: Soft Label Mechanism and Key-Extraction-Guided CoT-based...

CVE-2026-114493

LocalAI v4.6.0 through v4.11.0 contains a server-side request forgery vulnerability in core/gallery/gallery.go that validates only the initial gallery config URL. Attackers can submit a config URL through POST /models/apply that redirects to internal or cloud metadata addresses, receiving those...

CVE-2026-114493

LocalAI versions v4.6.0 through v4.11.0 are vulnerable to a Server-Side Request Forgery (SSRF) vulnerability within the core/gallery/gallery.go component. The root cause is that the application only validates the initial gallery configuration URL. An attacker can submit a configuration URL via PO...

CVE-2026-114493 LocalAI v4.6.0 through v4.11.0 SSRF Guard Bypass via HTTP Redirects

LocalAI v4.6.0 through v4.11.0 contains a server-side request forgery vulnerability in core/gallery/gallery.go that validates only the initial gallery config URL. Attackers can submit a config URL through POST /models/apply that redirects to internal or cloud metadata addresses, receiving those...

EUVD-2026-107421

LocalAI v4.6.0 through v4.11.0 contains a server-side request forgery vulnerability in core/gallery/gallery.go that validates only the initial gallery config URL. Attackers can submit a config URL through POST /models/apply that redirects to internal or cloud metadata addresses, receiving those...

CVE-2026-114493 LocalAI v4.6.0 through v4.11.0 SSRF Guard Bypass via HTTP Redirects

LocalAI v4.6.0 through v4.11.0 contains a server-side request forgery vulnerability in core/gallery/gallery.go that validates only the initial gallery config URL. Attackers can submit a config URL through POST /models/apply that redirects to internal or cloud metadata addresses, receiving those...

Learning-to-Detect

Learning to Detect Unknown Jailbreak Attacks in Large Vision-Language Models Official implementation of “Learning to Detect Unknown Jailbreak Attacks in Large Vision-Language Models.” This repository contains the data-processing, hidden-state extraction, classifier training, safety-pattern...

Insufficient Verification of Data Authenticity

Overview vllm is an A high-throughput and memory-efficient inference and serving engine for LLMs Affected versions of this package are vulnerable to Insufficient Verification of Data Authenticity via a forged client-supplied uuid field in multimodal content parts, which vLLM uses verbatim as the...

PT-2026-120345

A weakness has been identified in Tenstorrent tt-metal up to 0.80.0. The affected element is the function load hf state dict filtered of the file models/tt transformers/tt/load checkpoints.py. This manipulation of the argument weight map causes path traversal. The attack may be initiated remotely...

There's only so much code: the rise in vulnerability numbers

Frontier AI models like Mythos are rapidly mining decades of technical debt across ubiquitous legacy libraries, driving monthly Microsoft Patch Tuesday vulnerability advisories up more than tenfold reaching 1,415 this month. This volume is overwhelming open-source maintainers and putting pressure...

TNC-Defense

TNC-Defense TNC-Defense provides research code for detecting and detoxifying backdoors in text-to-image diffusion models. The repository currently includes detection pipelines for Stable Diffusion v1.4, Stable Diffusion v1.5, Stable Diffusion XL, and Stable Diffusion 3 Medium, together with...

rats-re

Rats! source reconstruction with local LLMs This repository is a work-in-progress reconstruction of the source code for RATS.EXE, the original Windows version of Rats! 1994 by Sean O'Connor. It builds a Win32 executable with Microsoft Visual C++ 4.1 under wibo and can be tested in DREAMM. The...