CVE-2026-31232

The CosyVoice project thru commit 6e01309e01bc93bbeb83bdd996b1182a81aaf11e 2025-30-21 contains an insecure deserialization vulnerability CWE-502 in its model loading process. When loading model files .pt from a user-specified directory via the --modeldir argument, the code uses torch.load without...

Iterative Audit Convergence in LLM-Managed Multi-Agent Systems: A Case Study in Prompt Engineering Quality Assurance

Prompt specifications for multi-agent large language model LLM systems carry data contracts and integration logic across many interdependent files but are rarely subjected to structured-inspection rigor. This paper reports a single-system empirical case study of iterative, agent-driven auditing...

PT-2026-40119

The CosyVoice project thru commit 6e01309e01bc93bbeb83bdd996b1182a81aaf11e 2025-30-21 contains an insecure deserialization vulnerability CWE-502 in its model loading process. When loading model files .pt from a user-specified directory via the --model dir argument, the code uses torch.load withou...

CVE-2026-31222

The snorkel library thru v0.10.0 contains an insecure deserialization vulnerability CWE-502 in the Trainer.load method of the Trainer class. The method loads model checkpoint files using torch.load without enabling the security-restrictive weightsonly=True parameter. This default behavior allows...

CVE-2026-31224

The snorkel library thru v0.10.0 contains an insecure deserialization vulnerability CWE-502 in the MultitaskClassifier.load method of the MultitaskClassifier class. The method loads model weight files using torch.load without enabling the security-restrictive weightsonly=True parameter. This...

CVE-2026-31232

CVE-2026-31232 affects the CosyVoice project; insecure deserialization (CWE-502) in model loading via --model_dir allows loading .pt files with pickle payloads. torch.load() is called without weights_only=True, enabling arbitrary Python object deserialization and remote code execution when a vict...

CVE-2026-31229

The ART (Adversarial Robustness Toolbox) package up to v1.20.1 contains an insecure deserialization vulnerability in its Kubeflow component’s model loading path. Loading model weights (e.g., model.pt) uses torch.load() without weights_only=True, allowing arbitrary Python object deserialization vi...

SkillSafetyBench: Evaluating Agent Safety under Skill-Facing Attack Surfaces

Reusable skills are becoming a common interface for extending large language model agents, packaging procedural guidance with access to files, tools, memory, and execution environments. However, this modularity introduces attack surfaces that are largely missed by existing safety evaluations: eve...

CVE-2026-31217

The loadmodel function in the neuralmagictraining.py script of the optimate project in commit a6d302f912b481c94370811af6b11402f51d377f 2024-07-21 allows arbitrary code execution. When a user supplies a directory path via the --model command-line argument, the function reads a module.py file from...

CVE-2026-31217

The CVE-2026-31217 entry concerns the optimate project’s neural_magic_training.py _load_model() function. If a user supplies a directory via --model, it reads module.py from that directory and executes its contents with Python's exec() without validation or sanitization. This enables an attacker ...

LLM 安全漏洞

LLM is a multi-model large language model command-line interaction tool developed by Simon Willison. Versions of LLM 0.27.1 and earlier contain security vulnerabilities. These vulnerabilities stem from the use of the --functions command-line parameter to directly execute unsafe code using the exe...

CVE-2026-31238

The Ludwig framework thru 0.10.4 is vulnerable to insecure deserialization CWE-502 in its model serving component. When starting a model server with the ludwig serve command, the framework loads model weight files using torch.load without enabling the security-restrictive weightsonly=True...

Reconstruction of Personally Identifiable Information from Supervised Finetuned Models

Supervised Finetuning SFT has become one of the primary methods for adapting a large language model LLM with extensive pre-trained knowledge to domain-specific, instruction-following tasks. SFT datasets, composed of instruction-response pairs, often include user-provided information that may...

ludwig 安全漏洞

Ludwig is an open-source declarative deep learning framework developed by Ludwig. Versions of Ludwig 0.10.4 and earlier contain security vulnerabilities. These vulnerabilities stem from the model service component using torch.load without enabling the weightsonly=True parameter when loading model...

VMware Spring AI 安全漏洞

VMware Spring AI is a development framework created by VMware Corporation in the Spring ecosystem, which integrates artificial intelligence and large language model capabilities. VMware Spring AI has a security vulnerability. This vulnerability allows malicious users to manipulate the behavior of...

PT-2026-40007

A malicious user could craft input that is stored in conversation memory and later interpreted by the model in an unintended way. Applications using the affected advisor with user-controlled input may be susceptible to manipulation of model behavior across conversation turns...

PT-2026-40058

The load model function in the neural magic training.py script of the optimate project in commit a6d302f912b481c94370811af6b11402f51d377f 2024-07-21 is vulnerable to insecure deserialization CWE-502. When a user provides a single model file path e.g., .pt or .pth via the --model command-line...

CTFusion: A CTF-Based Benchmark for LLM Agent Evaluation

Recent advances in Large Language Models LLMs have enabled agentic systems for complex, multi-step tasks; cybersecurity is emerging as a prominent application. To evaluate such agents, researchers widely adopt Capture The Flag CTF benchmarks. However, current CTF benchmarks reuse existing...

Adversarial Robustness Toolbox 安全漏洞

Adversarial Robustness Toolbox is an open-source machine learning security defense and evaluation tool developed by Trusted-AI. Versions of Adversarial Robustness Toolbox 1.20.1 and earlier contained security vulnerabilities. These vulnerabilities stemmed from the model loading function in the...

CVE-2026-31229

The Adversarial Robustness Toolbox ART thru 1.20.1 contains an insecure deserialization vulnerability CWE-502 in its Kubeflow component's model loading functionality. When loading model weights from a file e.g., model.pt during robustness evaluation, the code uses torch.load without the...