Security updates and support are provided for the latest and most recent AIStore releases. Keep deployments on the latest stable release.
| Version | Supported |
|---|---|
| Latest release | ✅ |
main branch |
✅ |
| Older versions | ❌ |
Do not open a public issue or pull request. Report vulnerabilities privately using one of these methods:
- NVIDIA Vulnerability Disclosure Program (preferred): Submit the official NVIDIA vulnerability report.
- GitHub private vulnerability reporting: Open this repository's Security tab and select Report a vulnerability.
- Email: Send vulnerability details to psirt@nvidia.com and copy aistore@nvidia.com for AIStore-specific coordination.
Include the affected version or branch, vulnerability type, reproduction steps, proof of concept if available, and expected impact. NVIDIA PSIRT will acknowledge, assess, and coordinate remediation and disclosure.
AIStore is a distributed object-storage service with native HTTP and S3-compatible APIs, proxy and target nodes, remote cloud backends, and command-line and SDK clients. Its main security boundaries are client-facing APIs, intra-cluster networks, storage hosts, and remote backends.
Repository Exposure Classification: Public. Basis: published in a public NVIDIA GitHub repository.
Service Exposure Classification: External / Regulated (high confidence). Basis: externally distributed production storage software and client APIs.
- Unauthorized data or administrative access: AIStore APIs manage objects, buckets, cluster membership, and batch operations. Missing or incorrect JWT, TLS, or network controls can expose data or privileged actions.
- Malicious or oversized input: Object, archive, S3, and batch APIs process caller-controlled names, metadata, and payloads. Crafted requests can target parsers or exhaust network, memory, CPU, or storage resources.
- Compromised cluster member: Proxies and targets exchange control, metadata, and object traffic. A compromised node or host can affect cluster integrity, confidentiality, or availability.
- Backend credential exposure: Cloud backends use external credentials and endpoints. Leaked or over-privileged credentials can expose remote data beyond the AIS cluster.
- Operators correctly configure TLS, JWT authentication and authorization, firewalls, and public versus intra-cluster networks.
- Cluster hosts, Kubernetes, storage devices, and the container runtime enforce their isolation and access controls.
- Cloud credentials and TLS private keys are protected and limited to the access required by the deployment.
- Only trusted administrators can perform destructive cluster operations or initialize ETL workloads.