AIKIDO-2026-376114
kedro-datasets is vulnerable to Deserialization of Untrusted Data
59
Medium Risk
This Affects:
kedro-datasetsTL;DR
kedro_datasets_experimental.pytorch.PyTorchDataset in kedro-datasets loads model files through torch.load without enforcing weights_only=True. User-supplied load_args are silently dropped and never reach torch.load, so catalog configuration cannot restrict deserialization. Loading a crafted model file from an untrusted source deserializes arbitrary pickle data and executes arbitrary code in the process running the pipeline. The fix enforces weights_only=True by default and forwards load_args to torch.load.
Who does this affect?
You are affected if you are using a version that falls within the vulnerable range and your project loads PyTorch model files from untrusted sources through kedro_datasets_experimental.pytorch.PyTorchDataset.
Background info
kedro-datasets is vulnerable to Deserialization of Untrusted Data in versions 5.0.0 - 9.4.0.
How to fix this
Upgrade the kedro-datasets library to the patch version.
Are You Affected?
Connect your repositories to instantly see whether vulnerable or malicious packages exist in your codebase.
Free. No credit card required.
I consent to receiving marketing communications based on Aikido’s Privacy Policy.
SOC 2Compliant
ISO 27001Compliant