Core 6 | Security spine
Secure AI Engineering: MLOps, LLMOps, and Supply Chain
Part of the Applied AI Security and Assurance curriculum, authored by Cameron Hopkin, CISSP, CEH, CHFI. Published as a reference, not an enrollable course.
Description
Real production security for AI systems. This is also where the curriculum is honest about tooling: PyTorch is the research and training standard, but TensorFlow's production ecosystem still leads a large share of enterprise deployment [4], so the material covers securing the stack that production actually runs, not only the one research prefers.
Outcomes
- Build data and model pipelines with integrity guarantees.
- Establish model and dependency provenance.
- Design monitoring and incident response for AI systems in production.
- Secure both research-native and production-native toolchains.
Modules
- Data pipelines and their integrity guarantees.
- Model and dependency supply chain: provenance, signing, and SBOM for models.
- Deployment and serving across research and production stacks.
- Monitoring, drift, and incident response for AI systems.
- Secrets, access, and isolation in AI infrastructure.
- Tamper detection end to end.
Signature lab
Build a pipeline with provenance and tamper detection across the whole path from data to deployed model.
Research thread
Practical integrity guarantees for a model supply chain.