Security of Machine Learning Systems
summary
Shows concrete attacks and engineering, not vague theory: live demos use published adversarial code to turn a golf ball image into a rifle with adversarial image perturbation (97% fooled), show patches that evade facial classifiers, sketches of NLP inputs that trigger unexpected outputs, and discuss steganography hiding malware in binaries. Covers MLOps realities — CI/CD with continuous training, model registries, and data poisoning risks — and launches an ML Top 10 draft and GitHub contributions process seeking reviewers.
AI-written summary · about
chapters (10)
- 0:00Welcome
- 1:03Presenter overview
- 4:18AI and machine learning fundamentals for beginners
- 15:26Key machine learning terminology explained
- 22:10ML in software systems
- 28:33ML security risks
- 33:05Adversarial attacks
- 39:45OAS ML Security Top 10
- 44:23Contributing to project
- 52:20Q&A
related talks
28mSecurity of Machine Learning
36mAdversarial Machine Learning: How Secure Is ML?
02h 13mUnderstanding Security Threats Against Machine/Deep Learning Applications
35mHacking AI Systems: How to (Still) Trick Artificial Intelligence
46mA Short Introduction to Hacking Artificial Intelligence
52mSleeping With One AI Open: An Introduction to Attacks Against...
20mMachine Learning Attacks – A New Era of Security Epidemic
39mSecure Machine Learning: Overcoming Algorithmic Bias, Adversarial ML in the Wild West of AI