Data poisoning can corrupt the information AI and machine-learning systems rely on, with consequences ranging from misleading search results to risks for creators, security teams, and critical decisions. Brian and Milou explore offensive and defensive uses of data poisoning, including ways artists try to protect their work, and discuss practical defenses such as validating data sources, monitoring unusual access and changes, sanitizing inputs, and testing response plans.
Chapters
00:00 Understanding Data Poisoning
03:50 Defensive Data Poisoning
09:50 Offensive and Defensive Strategies
11:50 Beyond AI: Polls, Studies and Data Brokers
16:45 Real-World Implications of Data Poisoning
20:50 Detection and Prevention Techniques
23:30 Detecting Anomalies
28:25 Data Sanitization
29:40 Enterprise Defense Strategies
32:30 Wrap-Up and Listener Feedback
No comments yet. Be the first to say something!