Explore a Google TechTalk on membership inference attacks (MIA) in machine learning, presented by Reza Shokri. Delve into the innovative RMIA algorithm, designed for efficient and robust detection of training data usage in ML models. Learn how this novel approach outperforms existing methods in accuracy and computational efficiency, even under limited resources. Discover the importance of stress testing MIA techniques for practical privacy risk assessment. Gain insights into the groundbreaking potential of RMIA for reliable data privacy analysis in machine learning applications.
Low Cost High Power Membership Inference Attacks
Overview
Syllabus
Low Cost High Power Membership Inference Attacks
Taught by
Google TechTalks
Related Courses
-
Membership Inference Attacks against Machine Learning Models
-
Machine Unlearning and Privacy Implications - Differential Privacy for ML
-
Secure Self-supervised Learning: Challenges and Solutions
-
Membership Inference Attacks from First Principles
-
Auditing Data Privacy for Machine Learning
-
Membership Inference Attacks against Adversarially Robust Deep Learning Models
