Large Language Model Security: Model Extraction Attacks Explained

Large Language Model Security: Model Extraction Attacks Explained Join Matt and Danny as they dive deep into the world of large language model security in a new video in their series. In this episode, they focus on model extraction attacks, a method by which attackers can steal valuable information by querying a model extensively. Understanding the working mechanism of such attacks, the defense strategies against them, and their detection difficulties are also discussed. Watch out for the next episode where 'jailbreaking' will be explained in detail. 00:00 Gangnam Style 00:05 Intro 00:25 What is a model extraction attack? 00:49 How do you steal models? 02:17 How can you defend against it? 03:38 What's next? 03:56 Outtakes