# CVE-2026-130638: AWS AutoGluon MultiModal path traversal in dataset tool

Source: https://junglewise.ai/threats/cve-2026-130638-path-traversal-in-the-object-detection-dataset-preparation-tool  
Publisher: Junglewise Threat Intelligence, https://junglewise.ai/threats  
Published: 2026-11-18. Updated: 2026-11-18  
License: CC BY 4.0 with a link back

CVE-2026-130638 · Severity: high · CVSS 6.8 · Published 2026-11-18

AutoGluon MultiModal is an open-source machine learning framework for training AI models. The object detection dataset preparation tool downloads public benchmark datasets but failed to validate archive contents, allowing an attacker on the network to substitute a malicious archive and write arbitrary files with the user's privileges. This could lead to code execution or system compromise, though the core training and inference APIs are unaffected.

Technologies: [Amazon AWS](https://junglewise.ai/threats/technologies/aws). Vendors: [Amazon](https://junglewise.ai/threats/vendors/amazon), [AWS](https://junglewise.ai/threats/vendors/aws).

## Quick answers

**Is CVE-2026-130638 exploited in the wild?** No. Junglewise Threat Intelligence has no report of CVE-2026-130638 being exploited in the wild.

**How severe is CVE-2026-130638?** High severity, with a CVSS score of 6.8, published 18 November 2026.

**Which products and versions are affected?** AWS AutoGluon MultiModal before 1.6.0.

**Is there a fix for CVE-2026-130638?** Yes: fixed in AutoGluon MultiModal 1.6.0. Until then: Do not run the prepare_detection_dataset utility. Instead, download datasets directly from their official sources over HTTPS and extract them with a tool that validates archive members, such as Python's tarfile module with filter="data". Junglewise finds every place you run it, drafts the upgrade and applies it once you approve.

**How is CVE-2026-130638 attacked?** It is attacked over the network or from the adjacent network, in a software library.

**Am I affected by CVE-2026-130638?** Check whether you run Amazon AWS or AWS AutoGluon MultiModal at an affected version. Junglewise does that check for you around the clock: connect your code, cloud and domains, and it matches new vulnerabilities like this one against what you actually run and tells you when one reaches you.

## How to fix

- Fixed in AutoGluon MultiModal 1.6.0

Do not run the prepare_detection_dataset utility. Instead, download datasets directly from their official sources over HTTPS and extract them with a tool that validates archive members, such as Python's tarfile module with filter="data".

## Technical details

A relative path traversal vulnerability exists in the prepare_detection_dataset utility, which downloads dataset archives over plaintext HTTP and extracts them without validating archive members. An adversary-in-the-middle attacker can substitute a crafted archive containing absolute paths or ".." directory references to write files outside the intended output directory. The vulnerability requires network-layer interception of unencrypted HTTP traffic and affects versions before 1.6.0.

## Affected products

- AWS AutoGluon MultiModal before 1.6.0

## Timeline

- 2026-11-18: disclosed

## References

- [https://aws.amazon.com/security/security-bulletins/rss/2026-188-aws/](https://aws.amazon.com/security/security-bulletins/rss/2026-188-aws/)

## Related threats

- [AWS SDK for PHP S3 Encryption Client GCM authentication tag downgrade](https://junglewise.ai/threats/cve-2026-132607-insufficient-verification-of-data-authenticity-in-the-decryption)
- [PHP multiple vulnerabilities in core and extensions](https://junglewise.ai/threats/php-foundation-security-releases-november-19-2026-b716b638)
- [Ruby heap-based buffer overflow in case conversion](https://junglewise.ai/threats/cve-2026-132825-ruby-before-3-3-14-3-4-13-and-4-0-10-has-a-heap-based-buffer)
- [NGINX heap buffer overflow in HTTP/2 flow control](https://junglewise.ai/threats/cve-2026-123007-this-issue-affects-nginx-plus-or-nginx-open-source)
- [OpenJDK Base64 Decoder Intrinsic Out-of-Bounds Write Issue](https://junglewise.ai/threats/cve-2026-114938-gcp-2026-082)
