DevSecOps for AI: Introducing MLSecOps for Software 2.0
summary
Presents an actionable MLSecOps framework that maps specific controls to pipeline stages: planning with risk registers and threat modeling, dataset versioning and poisoning checks, model provenance and model store integrity, secure packaging (avoid insecure pickle deserialization), adversarial training or differential privacy / federated learning, runtime defenses like rate limiting and mutual auth, and monitoring with drift and anomaly alerts. Claims prior penetration tests hit 100% success rate and cites an MLSEC AI hacking competition where top teams all bypassed fashion‑recognition models.
AI-written summary · about
chapters (12)
- 0:00MLSecOps for AI systems
- 3:46AI future of software
- 8:21Security threats to AI systems
- 13:41Shifting security left in AI
- 14:01MLSecOps framework overview
- 17:32Data collection risks
- 21:08Model building and dependencies
- 26:49Model validation challenges
- 31:04Model deployment threats
- 35:20Model monitoring and response
- 39:18AI pipeline security considerations
- 40:47Key takeaways and action
related talks
27mA Gentle Introduction to MLSecOps
15mApplying DevSecOps Lessons To MLSecOps
50mBuilding AI Security In: MLSecOps in Practice
26mRunning Secure AI/ML Experiment in a Unified MLOPS-DevSecOps Pipeline
53mSecurity of Machine Learning Systems
42mMLOps Under Attack: Threat Modeling Modern AI Systems
29mSecurity-Focused Chaos Engineering - the Lasso for AI Security T...
18mCommit Secure AI: Embedding Security in Code, Pipelines, and Production