From the course: Securing the Use of Generative AI in Your Organization
Join today to access over 24,700 courses taught by industry experts.
Model Inversion and Data Leakage
From the course: Securing the Use of Generative AI in Your Organization
Model Inversion and Data Leakage
All right. It's time for the world of model inversion and data leakage attacks in generative AI and large language models. These sneaky attacks pose serious threats to the security and integrity of AI models. Let's start with model inversion attacks. These attacks are all about prying open the security and confidentiality of AI models. Sneaky adversaries try to extract sensitive information from the model's outputs. Even without direct access to the underlying data, they exploit vulnerabilities in large language models that unintentionally leak information. Large language models often rely heavily on their training data, making them susceptible to reverse engineering tricks. The outputs of large language models can accidentally spill the beans about the training data, allowing attackers to piece together sensitive details hidden in the training data set. Model inversion attacks give attackers the power to extract private or proprietary information from the model's outputs, putting the…
Practice while you learn with exercise files
Download the files the instructor uses to teach the course. Follow along and learn by watching, listening and practicing.
Download courses and learn on the go
Watch courses on your mobile device without an internet connection. Download courses using your iOS or Android LinkedIn Learning app.
