# CVE-2026-79919: MaxKB sandbox bypass via dlopen and dlsym

Source: https://junglewise.ai/threats/cve-2026-79919-maxkb-is-an-open-source-ai-assistant-for-enterprise-prior-to  
Publisher: Junglewise Threat Intelligence, https://junglewise.ai/threats  
Published: 2026-09-21. Updated: 2026-09-26  
License: CC BY 4.0 with a link back

CVE-2026-79919 · Severity: medium · CVSS 6.3 · Published 2026-09-21

MaxKB is an open-source AI assistant platform that executes untrusted code in a sandboxed environment. Prior to version 2.10.6-lts, an authenticated workspace member could bypass the sandbox's syscall restrictions by using ctypes and unhooked dlsym calls to directly invoke glibc functions, allowing them to read/write files, execute processes, or access networks as the sandbox user.

Technologies: [1Panel MaxKB](https://junglewise.ai/threats/technologies/1panel-maxkb). Vendors: [1Panel](https://junglewise.ai/threats/vendors/1panel).

## Quick answers

**Is CVE-2026-79919 exploited in the wild?** No. Junglewise Threat Intelligence has no report of CVE-2026-79919 being exploited in the wild.

**How severe is CVE-2026-79919?** Medium severity, with a CVSS score of 6.3, published 21 September 2026.

**Which products and versions are affected?** 1Panel MaxKB before 2.10.6-lts.

**Is there a fix for CVE-2026-79919?** A patch was published on 2026-09-03. Junglewise finds every place you run it, drafts the upgrade and applies it once you approve.

**How is CVE-2026-79919 attacked?** It is attacked over the network, in a software library.

**Am I affected by CVE-2026-79919?** Check whether you run 1Panel MaxKB 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.

## Technical details

The vulnerability exploits a flaw in the LD_PRELOAD sandbox's dlopen call-stack heuristic: an attacker can invoke ctypes.CDLL from an importlib.abc.MetaPathFinder callback so the dlopen inspection sees a Python import frame, then use unhooked dlsym with RTLD_NEXT to resolve glibc's real syscall function and bypass the sandbox blacklist. This requires authentication as a workspace member. The vulnerability is patched in version 2.10.6-lts.

## Affected products

- 1Panel MaxKB before 2.10.6-lts

## Timeline

- 2026-09-21: disclosed
- 2026-09-03: patched

## References

- [https://github.com/1Panel-dev/MaxKB/commit/d4bd22af8ba14a9dea317f9034332fec4e964697](https://github.com/1Panel-dev/MaxKB/commit/d4bd22af8ba14a9dea317f9034332fec4e964697)
- [https://github.com/1Panel-dev/MaxKB/releases/tag/v2.10.6-lts](https://github.com/1Panel-dev/MaxKB/releases/tag/v2.10.6-lts)

## Related threats

- [MaxKB ToolExecutor sandbox bypass via fexecve](https://junglewise.ai/threats/cve-2026-79918-maxkb-is-an-open-source-ai-assistant-for-enterprise-prior-to)
- [MaxKB chat share-link endpoint missing owner verification](https://junglewise.ai/threats/cve-2026-79917-maxkb-is-an-open-source-ai-assistant-for-enterprise-in-2-7-0)
- [MaxKB credential injection in AWS Bedrock configuration](https://junglewise.ai/threats/cve-2026-79916-maxkb-is-an-open-source-ai-assistant-for-enterprise-prior-to-2-10)
- [MaxKB missing per-tool authorization in agent and workflow dispatch](https://junglewise.ai/threats/cve-2026-77516-maxkb-is-an-open-source-ai-assistant-for-enterprise-from-version)