Threat Modeling in the Age of AI
summary
Claims a usable OWASP AI Exchange threat framework and a published ML SEC Ops architecture and shows how to apply them to concrete attacks like data poisoning and indirect prompt injection. Describes practical tooling and practices: Jupyter notebooks, gated repos, continuous data validation and monitoring, MLops pipelines. Gives a sobering number: a pregnancy model with 90% accuracy would misserve about 750,000 people so it can’t be ethically deployed. Emphasizes data provenance, access controls and operationalization as the things that actually change outcomes.
AI-written summary · about
chapters (9)
- 0:00Introduction
- 1:45Why threat modeling matters now
- 4:59What threat modeling is
- 8:32AI changes threat landscape
- 12:06AIML Ops security
- 20:57Three AI threat modeling questions
- 28:04Mapping AIML attack surface
- 35:14Resources and closing
- 35:57Q&A
related talks
25mThreat Modeling for AI Apps with Attacks as Code
18mCommit Secure AI: Embedding Security in Code, Pipelines, and Production
41mThreat Modelling for ML/AI systems
50mBuilding AI Security In: MLSecOps in Practice
01h 01mSecure AI: Integrating OWASP Principles in Machine Learning Model Development
49mThe Intersection of AI and Cybersecurity
55mAttacking And Protecting Artificial Intelligence
45mHardening AI/ML Systems - The Next Frontier of Cybersecurity