Database/AI/ML frameworks & serving
NVIDIA NemoClaw: deployment process fails to validate certificates properly
Impact
The deployment path accepts certificates it should reject, so anyone able to sit on the connection it makes can answer it instead of the intended endpoint. NVIDIA lists information disclosure, data tampering, code execution and privilege escalation as consequences, which is what you get when a deployment channel can be substituted: credentials and configuration flow to the wrong peer, and what comes back is trusted. Deployments typically run across many nodes from the same automation, so an intercepted deployment is a fleet-wide event rather than a single-node one.
Who can reach it
An attacker positioned to intercept the deployment connection (CVSS AV:N/PR:N, high complexity - the position is the hard part). No authentication to the product is required.
What to do
NVIDIA product-security bulletin 5872 carries the fixed version; the CVE record does not state one. Until it is applied, run deployments only over network paths you control end to end and re-deploy from a patched build afterwards, since anything deployed through the flawed path should be considered of unverified provenance.
References
Related entries
- NVIDIA NemoClaw: weak authentication in the remote-access helper workflowCVE-2026-65098 · NVIDIA NemoClaw for Linux (remote-access helper workflow)High
- NVIDIA NemoClaw: inference service comes up without authentication, reachable from the adjacent networkCVE-2026-65105 · NVIDIA NemoClaw for Linux (inference server setup)High
- ClearML client SDK: Deserialization of untrusted dataCVE-2024-24590 · ClearML client SDKHigh
- ClearML client SDK: Path traversal — a malicious dataset writes arbitrary files on the consumerCVE-2024-24591 · ClearML client SDKHigh
- Keras (`utils.get_file`): Path traversal in tar extraction in 3.11.3 (incomplete fix)CVE-2025-12638 · Keras (`utils.get_file`)High
- vLLM (multi-node ZeroMQ): Secondary vLLM host trusts unauthenticated ZeroMQ messagesCVE-2025-30165 · vLLM (multi-node ZeroMQ)High
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.