CVE-2026-65918 PyTorch torchvision GIF Decoder Out-of-bounds Heap Read

PyTorch torchvision through 0.28.0, fixed in commit 4e05dc2, contains an out-of-bounds heap read vulnerability in the GIF decoder's readfromtensor callback that passes unclamped length to memcpy. Attackers can supply malicious or truncated GIF files to cause denial of service via segmentation fau...

CVE-2026-65918

CVE-2026-65918 affects PyTorch torchvision up to version 0.28.0. The GIF decoder’s read_from_tensor callback performs an out-of-bounds heap read by passing an unclamped length to memcpy. Attackers can supply malicious or truncated GIF files to trigger a denial of service (segmentation fault) or d...

CVE-2026-65918 PyTorch torchvision GIF Decoder Out-of-bounds Heap Read

PyTorch torchvision through 0.28.0, fixed in commit 4e05dc2, contains an out-of-bounds heap read vulnerability in the GIF decoder's readfromtensor callback that passes unclamped length to memcpy. Attackers can supply malicious or truncated GIF files to cause denial of service via segmentation fau...

CVE-2026-12484

A flaw was found in Keras, specifically within the keras.layers.TorchModuleWrapper.fromconfig method. This vulnerability allows an attacker to perform unsafe deserialization of PyTorch pickle data. By processing untrusted Keras layer configurations, a local attacker could exploit this flaw, leadi...

Linux Distros Unpatched Vulnerability : CVE-2026-12484

The Linux/Unix host has one or more packages installed that are impacted by a vulnerability without a vendor supplied patch available. - A vulnerability in keras-team/keras version 3.15.0 allows unsafe deserialization of attacker-controlled PyTorch pickle data through the public...

GHSA-V2W2-W228-C444 Keras: TorchModuleWrapper can deserialize unsafe PyTorch pickle data

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 ...

DEBIAN-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-2026-12484

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 ...

EUVD-2026-45872

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-2026-12484

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-2026-12484 Unsafe Deserialization in keras.layers.TorchModuleWrapper.from_config

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

A flaw was found in PyTorch Lightning. This vulnerability allows attackers to achieve remote code execution by crafting malicious checkpoint files. These files exploit a weakness in the loadstate function, which imports and executes attacker-controlled module names from checkpoint instantiator...

Use of Externally-Controlled Input to Select Classes or Code ('Unsafe Reflection')

Overview pytorch-lightning is a lightweight PyTorch wrapper for ML researchers. Scale your models. Write less boilerplate. Affected versions of this package are vulnerable to Use of Externally-Controlled Input to Select Classes or Code 'Unsafe Reflection' via the loadstate function that imports a...

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...

CVE-2026-58659

Summary: CVE-2026-58659 affects PyTorch Lightning up to version 2.6.5. The vulnerability lies in the _load_state function, which can import and execute attacker-controlled module names from the checkpoint _instantiator hyperparameters, enabling remote code execution when LightningModule.load_from...

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...

CVE-2026-58659 PyTorch Lightning Arbitrary Code Execution via _instantiator Hyperparameter

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...

CVE-2026-58659 PyTorch Lightning Arbitrary Code Execution via _instantiator Hyperparameter

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...

PT-2026-60248

Name of the Vulnerable Software and Affected Versions PyTorch Lightning versions prior to 2.6.6 Description A remote code execution issue exists in the load state function. The flaw allows the import and execution of attacker-controlled module names from checkpoint instantiator hyperparameters. A...