Securing the web with AI
summary
Demos use real tools and numbers: anomaly detection costs about $0.15 per 1,000 checks, 800 million people faced malware in 2018 and nearly half a billion passwords are pwned. Implementations include Azure Cognitive Services for face and voice 2FA, a haveibeenpwned k‑anonymity check for passwords in .NET and JavaScript, in‑browser object and webcam analysis with TensorFlow.js coco‑ssd, and a Chrome extension (Fake Finder) that flags photoshopped images via compression and ML. Practical limits and spoofing risks are demonstrated.
AI-written summary · about
chapters (28)
- 0:00Introduction
- 1:15The web is broken
- 3:52Security is a right
- 9:33Educating users
- 12:42Password checks
- 16:36Multifactor authentication
- 25:12Pwned
- 25:28Solutions
- 26:45WebOffN
- 28:20Touch sensor
- 28:45Tensorflow
- 29:31Cocoa SSD
- 30:02Skit
- 30:27Demo
- 31:51Room key
- 32:21Webcam
- 33:26Cool stuff
- 33:44Fake news
- 34:03Image compression analysis
- 36:52Photo forensics
- 38:15Image forensics with AI
- 39:41Image forensics with JSON
- 41:09Chrome Extension
- 41:31Context Menu
- 42:00Fake Images
- 43:02Takeaways
- 43:52API and JavaScript
- 44:56Questions
related talks
11mHow to Navigate Emerging AI-Driven Security
20mSecurity in the Age of AI: Response and Strategy
37mFundamentals of AI safety and security
52mKnock Knock: Understanding Who is Using Your Web Applications
40mWeb Attacks and Fraud: Lessons Learned from the Deep End of the Pool
01h 03mTen Key Steps for Enhanced Web App Security
27mCybersecurity During Dark Times
57mAI as the New Hacker Frontier