CVE-2026-12484

A vulnerability in keras-team/keras version 3.15.0 allows unsafe deserialization of attacker-controlled PyTorch pickle data through the public keras.layers.TorchModuleWrapper.fromconfig method. This method invokes torch.load..., weightsonly=False without requiring an explicit unsafe opt-in, such ...

CVE-2026-58659

PyTorch Lightning through 2.6.5, fixed in commit d710d68, contains a remote code execution vulnerability in the loadstate function that imports and executes attacker-controlled module names from checkpoint instantiator hyperparameters. Attackers can craft malicious checkpoint files that bypass...

EUVD-2026-44751

PyTorch Lightning through 2.6.5, fixed in commit d710d68, contains a remote code execution vulnerability in the loadstate function that imports and executes attacker-controlled module names from checkpoint instantiator hyperparameters. Attackers can craft malicious checkpoint files that bypass...

PYSEC-2026-1856 PyTorch Vulnerable to Remote Code Execution via Untrusted Checkpoint Files

Summary 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 arbitrary code execution. Vulnerability Details The weightsonly=True unpickler...

PYSEC-2026-967 TensorFlow has Segfault in Bincount with XLA

Impact When running with XLA, tf.rawops.Bincount segfaults when given a parameter weights that is neither the same shape as parameter arr nor a length-0 tensor. python import tensorflow as tf func = tf.rawops.Bincount para='arr': 6, 'size': 804, 'weights': 52, 351 @tf.functionjitcompile=True def...

PYSEC-2026-1043 TensorFlow vulnerable to `CHECK` fail in `DenseBincount`

Impact DenseBincount assumes its input tensor weights to either have the same shape as its input tensor input or to be length-0. A different weights shape will trigger a CHECK fail that can be used to trigger a denial of service attack. python import tensorflow as tf binaryoutput = True input =...

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...

PYSEC-2026-405 Ludwig framework is vulnerable to insecure deserialization in its model serving component

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...

CVE-2026-47155 vLLM: Artifact Pin Decay in vLLM allows pinned deployments to load unpinned code, weights, and processors

vLLM is an inference and serving engine for large language models LLMs. Prior to 0.22.0, vLLM's revision pinning controls do not consistently apply to all artifacts loaded for a model. A deployment that supplies --revision or --code-revision can still load dynamic code, GGUF files, image...

CVE-2026-47155

CVE-2026-47155 affects vLLM prior to 0.22.0. Description: revision pinning controls do not consistently apply to all artifacts loaded for a model, enabling loading of dynamic code, GGUF files, image processors, retrieval side weights, or same-repository subfolder weights/config from an unpinned/d...

CVE-2026-47155 vLLM: Artifact Pin Decay in vLLM allows pinned deployments to load unpinned code, weights, and processors

vLLM is an inference and serving engine for large language models LLMs. Prior to 0.22.0, vLLM's revision pinning controls do not consistently apply to all artifacts loaded for a model. A deployment that supplies --revision or --code-revision can still load dynamic code, GGUF files, image...

GHSA-3WW4-5JV9-J5GM vLLM's Artifact Pin Decay allows pinned deployments to load unpinned code, weights, and processors

Summary vLLM's revision pinning controls do not consistently apply to all artifacts loaded for a model. A deployment that supplies --revision or --code-revision can still load dynamic code, GGUF files, image processors, retrieval side weights, or same-repository subfolder weights/config from an...

vLLM's Artifact Pin Decay allows pinned deployments to load unpinned code, weights, and processors

Summary vLLM's revision pinning controls do not consistently apply to all artifacts loaded for a model. A deployment that supplies --revision or --code-revision can still load dynamic code, GGUF files, image processors, retrieval side weights, or same-repository subfolder weights/config from an...

PT-2026-48537

Name of the Vulnerable Software and Affected Versions vLLM versions prior to 0.22.0 Description vLLM is an inference and serving engine for large language models. The software contains a supply-chain integrity issue where revision pinning controls are not consistently applied to all artifacts...

CVE-2026-31229

A flaw was found in the Adversarial Robustness Toolbox ART, specifically within its Kubeflow component. This vulnerability, categorized as insecure deserialization CWE-502, occurs when loading model weights without proper security restrictions. A remote attacker can exploit this by uploading a...

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-2026-31239

The mamba language model framework thru 2.2.6 is vulnerable to insecure deserialization CWE-502 when loading pre-trained models from HuggingFace Hub. The MambaLMHeadModel.frompretrained method uses torch.load to load the pytorchmodel.bin weight file without enabling the security-restrictive...

CRESS: Quantifying Vulnerabilities of Attack Scenarios in Hardware Reverse Engineering

The safety, security, and reliability of microelectronic systems depend on a trustworthy, secured supply chain and design flow. Globally distributed supply chains or unintentional design weaknesses leave the door open for attacks on the hardware level. These scenarios encompass counterfeiting,...

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-2026-31219

The loadmodel function in the neuralmagictraining.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 argumen...