# CVE-2026-112073: Tencent ncnn MatMul out-of-bounds read

Source: https://junglewise.ai/threats/cve-2026-112073-a-vulnerability-was-identified-in-tencent-ncnn-up-to  
Publisher: Junglewise Threat Intelligence, https://junglewise.ai/threats  
Published: 2026-10-19. Updated: 2026-10-19  
License: CC BY 4.0 with a link back

CVE-2026-112073 · Severity: medium · CVSS 6.3 · Published 2026-10-19

Tencent ncnn is a neural network inference framework for mobile platforms. A bug in the MatMul layer fails to validate consistency between input matrix dimensions, allowing an attacker to craft a malicious model that triggers an out-of-bounds memory read during matrix operations. This could crash the application or potentially leak sensitive data.

Technologies: [Tencent Ncnn](https://junglewise.ai/threats/technologies/ncnn). Vendors: [Tencent](https://junglewise.ai/threats/vendors/tencent).

## Quick answers

**Is CVE-2026-112073 exploited in the wild?** No. Junglewise Threat Intelligence has no report of CVE-2026-112073 being exploited in the wild.

**How severe is CVE-2026-112073?** Medium severity, with a CVSS score of 6.3, published 19 October 2026.

**Which products and versions are affected?** Tencent Ncnn up to commit 44b02dabdcf97499281a4692ebd65bad73dcf8b6.

**Is there a fix for CVE-2026-112073?** No fixed version is recorded yet. Check the vendor's advisory in the references below. If you run it, Junglewise drafts a mitigation that fits your setup and applies it once you approve, with senior security engineers behind it.

**How is CVE-2026-112073 attacked?** It is attacked over the network, in a software library.

**Am I affected by CVE-2026-112073?** Check whether you run Tencent Ncnn 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 MatMul layer in src/layer/matmul.cpp sets the inner dimension K from input A's width without validating it matches input B's width. When A.w exceeds B.w, the gemm_x86_avx.cpp transpose_pack_B_tile function reads past B's allocated buffer boundary. The vulnerability is triggered remotely by providing a crafted model file with mismatched input dimensions; no authentication or user interaction required.

## Affected products

- Tencent ncnn up to commit 44b02dabdcf97499281a4692ebd65bad73dcf8b6

## Timeline

- 2026-10-01: disclosed: Issue #7034 reported on GitHub
- 2026-10-19: advisory: CVE-2026-112073 published

## References

- [https://github.com/Tencent/ncnn/](https://github.com/Tencent/ncnn/)
- [https://github.com/Tencent/ncnn/issues/7034](https://github.com/Tencent/ncnn/issues/7034)

## Related threats

- [Tencent ncnn integer overflow in yuv420sp2rgb](https://junglewise.ai/threats/cve-2026-112074-a-security-flaw-has-been-discovered-in-tencent-ncnn-up-to)
- [Tencent ncnn heap out-of-bounds write in ParamDict::load_param](https://junglewise.ai/threats/cve-2026-50144-tencent-ncnn-heap-out-of-bounds-write-in-paramdict-load-param)