Megatron-Bridge: code injection via malicious input in the data merging and data shuffling tutorials
Impact
A local user with low privileges who can supply input to the data merging or data shuffling tutorial scripts gets code execution, privilege escalation, information disclosure and data tampering (CVSS 7.8, AV:L/AC:L/PR:L/UI:N/C:H/I:H/A:H). NVIDIA split the same tutorial code-injection class across 2 ids, one per affected tutorial. Note the earlier database wording of 'unsafe pickle deserialization' was wrong: the vendor classifies both as CWE-94 code injection.
Who can reach it
Malicious checkpoint
What to do
Upgrade Megatron-Bridge to 0.2.2 or later (all prior versions affected) and rebuild any training images that embed it; one upgrade closes both ids.
Also covers 1 CVE
The vendor assigned a separate id to each affected code path. They share this advisory, this score and this fix, so they are one entry here.
References
Related entries
- Megatron-Bridge: RCE via unsafe deserialization of untrusted model and config artifactsCVE-2026-24240 · Megatron-BridgeHigh
- Megatron-Bridge: SSRF in file operationsCVE-2026-24242 · Megatron-BridgeHigh
- Megatron-Bridge: Validation bypass via incorrect type comparisonCVE-2026-24246 · Megatron-BridgeHigh
- Megatron-Bridge: RCE via unsafe evaluation of loaded configCVE-2026-24249 · Megatron-BridgeHigh
- NeMo Framework: Unsafe deserialization of untrusted model/config data allows arbitrary code executionCVE-2025-33241 · NeMo FrameworkHigh
- NeMo Framework: Local command injection via unsanitized input passed to shell executionCVE-2025-33246 · NeMo FrameworkHigh
This entry is curated: imported from vendor advisories with machine assistance, not yet individually verified. Confirm against your vendor's advisory before acting, and report anything wrong.