AI Deception: Fooling (Artificial) Intelligence Is Easier than You Think
summary
Describes practical security weaknesses in AI systems (examples include the Dolphin attack and adversarial inputs that mislead self-driving cars) and treats AI failures as exploitable edge cases. Covers the cyber-lifecycle for AI, methods to assess and mitigate AI-related risks, and guidance for securely integrating machine learning into adversarial environments.
AI-written summary · about
chapters (13)
- 0:00Intro
- 3:11Security and the Hype-Cycle
- 4:26The Australian Challenge
- 6:13Adversarial Thinking
- 7:28Blind Spots
- 8:20Getting Hit by an Ostrich Adversarial Examples
- 9:34Dolphin Attack
- 10:28Define Al Deception
- 11:35Al Deception Threat
- 14:51Are We Better? HI Deception
- 16:06Risk Mitigation
- 18:48Harnessing Al for Cyber Security Attack Detection in Imperva Counter Breach
- 19:49Summary
related talks
34mHow to Trick with AI
35mHacking AI Systems: How to (Still) Trick Artificial Intelligence
12mQuick Look: No Exemptions for AI: New Technology = New Attack Surface
46mA Short Introduction to Hacking Artificial Intelligence
41mAI and cyber, or why your vacuum cleaner (still)...
22mAdversarial AI: Why it Matters Now
47mWhat AI Security means for you as a developer: Time-focused approach
44mHacking AI - How to Survive the AI Uprising