ML pipeline security audit

Threat model and pentest of your ML/LLM stack: data → training → deployment → operations.

• OWASP LLM Top 10 and MITRE ATLAS: vectors, checks, PoC on a test environment;
• Data poisoning and backdoor audit of datasets and pretrained weights;
• Supply chain integrity: SLSA level, artifact signatures (cosign/Sigstore), SBOM;
• Inference isolation in K8s: eBPF monitoring, network policies, secrets management;
• Report with prioritized findings and a mitigation plan.