CVE-2026-130638 - Path traversal in the object detection dataset preparation tool in AutoGluon MultiModal
Bulletin ID: 2026-188-AWS
Scope: AWS
Content Type: Important (requires attention)
Publication Date: 11/18/2026 10:00 AM PST
Description:
AutoGluon is an open source machine learning framework maintained by AWS.
We identified CVE-2026-130638, an issue in the object detection dataset preparation tool in AutoGluon MultiModal before 1.6.0. The tool downloaded dataset archives over plaintext HTTP and extracted them without validating the archive members. An actor with an on-path network position could substitute a crafted archive and write files outside the intended output directory with the privileges of the user running the tool. This tool is an optional command-line utility for downloading public benchmark datasets. The AutoGluon Multimodal training and inference APIs, including object detection with MultiModalPredictor, do not use it and are not affected.
Impacted versions: < 1.6.0
Resolution:
This issue has been addressed in autogluon.multimodal version 1.6.0. We recommend upgrading to the latest version and ensuring any forked or derivative code is patched to incorporate the new fixes.
Workarounds:
If you cannot upgrade, do not run the prepare_detection_dataset utility. Instead, download the datasets directly from their official sources over HTTPS and extract them with a tool that rejects absolute paths and .. members, such as Python's tarfile with filter="data".
References:
Acknowledgement:
We would like to thank sondt99 for collaborating on this issue through the coordinated vulnerability disclosure process.
Please email aws-security@amazon.com with any s ecurity questions or concerns.