AI Impact on Cybersecurity and Privacy
summary
Claims that GitHub Copilot output contains 40–45% vulnerable code and that AI-trained network models reach >99% accuracy after a few weeks guide the talk. Demonstrations show two products: Glog drops static-analysis alerts from 456 to 88 by adding app context and returns field-level remediation (sometimes five-line fixes), and Impressic trains models, deploys them to sensors within five minutes, and turned a 12,500‑finding, 25,000‑hour cleanup into four man‑days.
AI-written summary · about
chapters (23)
- 0:00Event introduction
- 2:07AI and cybersecurity overview
- 4:25Cybercrime costs and AI trends
- 6:11AI definitions and security apps
- 8:58AI-driven cyberattacks and threats
- 11:50Malware evolution and AI misuse
- 13:18AI model vulnerabilities and data poisoning
- 14:23Cybercrime weaponization and risk management
- 16:14False positives in security tools
- 18:45AI in network threat detection
- 22:21Software security testing challenges
- 24:49AI for security automation
- 27:37Live demo
- 28:19Machine learning vs AI
- 29:02Impressic network security dashboard
- 31:52Security analyst in the loop
- 33:38Training AI models for threats
- 35:44Glog project background
- 37:11Static analysis tools and false positives
- 38:15Glog results vs SAST tools
- 41:58Context-aware security remediation
- 45:14Project status and future
- 47:03Q&A
related talks
01h 02mHow AI is impacting cybersecurity. Does it?
30mHuman vs AI: How to Ship Secure Code
53mHow AI is impacting cybersecurity. Does it?
58mSecuring Code in the Age of AI
38mCode Security Reinvented: Navigating the era of AI
41mCode Security Reinvented Navigating the era of AI
10mAI Coding & Cybersecurity — Why AI coding tools aren’t ready for production yet
01h 24mOffensive Cybersecurity by AI Promises and Pitfalls