Lessons Learned Developing an AI Cyber Immune System: The Good, Bad & Ugly
summary
An industrial immune system called digital ghost uses a high‑fidelity digital twin plus about 40 machine learning algorithms to monitor physics (voltage, frequency, temperature) rather than just network signatures, detect and localize sensor/actuator manipulation, and neutralize attacks in OT/ICS; live tests on a large combined‑cycle power plant show detection/localization with ~99% accuracy. It emphasizes humans + curated training data, warns of adversarial poisoning, deep‑fake/social engineering, black‑box limits, 350,000 new malware samples/day and ~2 million unfilled cyber roles, and calls for explainable and humble AI.
AI-written summary · about
chapters (7)
related talks
16mIntelligent by design: Can cybersecurity cope with AI?
44mAI Meets Cybersecurity: Boosting Speed Without Losing Control
11mAI-Powered Cybersecurity in the Age of Advanced Threats
27m"Cybertelligence": How AI is changing the face of cybersecurity
49mGenAI and Cybersecurity: Safeguarding the Digital Age
01h 24mOffensive Cybersecurity by AI Promises and Pitfalls
49mThe Intersection of AI and Cybersecurity
01h 19mAI vs AI: Cyber Defence and Managed Resilience