CVE-2026-31252

CosyVoice thru commit 6e01309e01bc93bbeb83bdd996b1182a81aaf11e 2025-30-21 contains an insecure deserialization vulnerability CWE-502 in its model loading component. The framework uses torch.load to load model weight files e.g., llm.pt, flow.pt, hift.pt without enabling the security-restrictive...

CVE-2026-31251

CVE-2026-31251 affects CosyVoice’s gRPC server component. During startup, the server loads the speech synthesis model from a user-specified directory via torch.load() without enabling the weights_only=True security parameter, enabling the pickle-based deserialization of arbitrary Python objects. ...

CVE-2026-31249

CosyVoice contains an insecure deserialization vulnerability (CWE-502) in its data processing tool make_parquet_list.py. The script loads PyTorch .pt files (utterance embeddings, speaker embeddings, speech tokens) with torch.load() without enabling weights_only=True, allowing the deserialization ...

CVE-2026-31250

CosyVoice thru commit 6e01309e01bc93bbeb83bdd996b1182a81aaf11e 2025-30-21 contains an insecure deserialization vulnerability CWE-502 in its averagemodel.py model averaging tool. The script loads PyTorch checkpoint files epoch.pt for model averaging using torch.load without enabling the...

CVE-2026-31252

CosyVoice thru commit 6e01309e01bc93bbeb83bdd996b1182a81aaf11e 2025-30-21 contains an insecure deserialization vulnerability CWE-502 in its model loading component. The framework uses torch.load to load model weight files e.g., llm.pt, flow.pt, hift.pt without enabling the security-restrictive...

CosyVoice 安全漏洞

CosyVoice is an open-source voice generation and AI voice cloning platform developed by FunAudioLLM. CosyVoice has a security vulnerability, which stems from the gRPC server component using torch.load to load the voice synthesis model without enabling the weights-only=True security parameter. Thi...

CVE-2026-31249

CosyVoice thru commit 6e01309e01bc93bbeb83bdd996b1182a81aaf11e 2025-30-21 contains an insecure deserialization vulnerability CWE-502 in its makeparquetlist.py data processing tool. The script loads PyTorch .pt files utterance embeddings, speaker embeddings, speech tokens using torch.load without...

CVE-2026-31249

CosyVoice thru commit 6e01309e01bc93bbeb83bdd996b1182a81aaf11e 2025-30-21 contains an insecure deserialization vulnerability CWE-502 in its makeparquetlist.py data processing tool. The script loads PyTorch .pt files utterance embeddings, speaker embeddings, speech tokens using torch.load without...

FlashAttention 安全漏洞

FlashAttention is an efficient and memory-efficient attention mechanism implementation tool open-sourced by Dao AI Lab. There is a security vulnerability in FlashAttention, which stems from the checkpoint loading mechanism using torch.load to load checkpoint files without enabling the...

openSUSE 16 Security Update : iproute2 (openSUSE-SU-2026:20696-1)

The remote openSUSE 16 host has packages installed that are affected by a vulnerability as referenced in the openSUSE- SU-2026:20696-1 advisory. Security issues fixed: - CVE-2024-58251: terminal lock up via ANSI terminal escape sequence set in argv0 bsc1254324. Other updates and bugfixes: - Fix...

CVE-2026-1839

A flaw was found in HuggingFace Transformers. A remote attacker can exploit this vulnerability by supplying a specially crafted checkpoint file e.g., rngstate.pth. The loadrngstate method in the Trainer class loads this file using torch.load without proper validation, specifically missing the...

Hugging Face Transformers 安全漏洞

Hugging Face Transformers is an open-source framework developed by Hugging Face for defining state-of-the-art machine learning models. It covers text, visual, audio, and multi-modal models, and can be used for both inference and training. There is a security vulnerability in Hugging Face...

Deserialization of Untrusted Data

Overview torch is a Tensors and Dynamic neural networks in Python with strong GPU acceleration Affected versions of this package are vulnerable to Deserialization of Untrusted Data in the .pt2 Loading Handler. An attacker can execute arbitrary code or alter application behavior by providing...

Activation Surgery: Jailbreaking White-Box LLMs without Touching the Prompt

Most jailbreak techniques for Large Language Models LLMs primarily rely on prompt modifications, including paraphrasing, obfuscation, or conversational strategies. Meanwhile, abliteration techniques also known as targeted ablations of internal components have been used to study and explain LLM...

model.weights.h5: h5py.ExternalLink at Group level silently followed during load_model(), bypassing CVE-2025-9905 fix — information disclosure from arbitrary HDF5 files

Keras 3.x introduced a fix for CVE-2025-9905 by checking dataset.external in H5IOStore.verifydataset. This check blocks datasets whose raw bytes are stored in external files via the HDF5 "External Data Storage" mechanism. However, HDF5 supports a second, unrelated external-reference mechanism:...

Incomplete Fix for CVE-2026-1669: HDF5 External Storage File Disclosure in Legacy H5 Loading

Description Keras 3 patched CVE-2026-1669 HDF5 External Storage File Disclosure in the new .keras and .weights.h5 loading paths by adding verifydataset to check for dataset.external in H5IOStore. However, the legacy .h5 loading path keras/src/legacy/saving/legacyh5format.py was not patched. This...

Deserialization of Untrusted Data

Overview nemo-toolkit is a NeMo - a toolkit for Conversational AI Affected versions of this package are vulnerable to Deserialization of Untrusted Data via the model loading process with weightsonly=False. An attacker can execute arbitrary code, escalate privileges, disclose sensitive information...

Vulnerabilities in Partial TEE-Shielded LLM Inference with Precomputed Noise

The deployment of large language models LLMs on third-party devices requires new ways to protect model intellectual property. While Trusted Execution Environments TEEs offer a promising solution, their performance limits can lead to a critical compromise: using a precomputed, static secret basis ...

BIT-PYTORCH-2026-24747 PyTorch Vulnerable to Remote Code Execution via Untrusted Checkpoint Files

PyTorch is a Python package that provides tensor computation. Prior to version 2.10.0, a vulnerability in PyTorch's weightsonly unpickler allows an attacker to craft a malicious checkpoint file .pth that, when loaded with torch.load..., weightsonly=True, can corrupt memory and potentially lead to...

CVE-2026-24747

A flaw was found in PyTorch, a Python package for tensor computation. A remote attacker could craft a malicious checkpoint file, which, when loaded using the weightsonly unpickler, could lead to memory corruption. This vulnerability may enable an attacker to achieve arbitrary code execution on th...