Explore privacy-preserving machine learning techniques in this 47-minute RSA Conference talk. Delve into the world of AI and privacy, learning how to combine these seemingly conflicting concepts without compromise. Discover emerging techniques that unlock AI's power while maintaining data privacy and confidentiality. Examine the higher computation and storage requirements of these methods, and understand recent research advancements in performance and usability. Investigate topics such as federated learning, trusted execution environments, and homomorphic encryption. Gain insights into real-world applications, including brain tumor segmentation challenges and the benefits of increased data availability. Analyze the mechanics, caveats, and architectures of various privacy-preserving approaches. Conclude with a discussion on future developments and potential applications in adverse settings.
Protect Privacy in a Data-Driven World - Privacy-Preserving Machine Learning
Overview
Syllabus
Intro
DataDriven World
Current Approaches to Privacy
Trust
Machine Learning
PrivacyPreserving Machine Learning
Use Case
Outline
Unreasonable Effectiveness
Data Silo
Mechanics of Federated Learning
Caveats
Trusted Execution Environments
Federated Learning Architecture
Integrity and attestation features
Data science caveat
Brain tumor segmentation challenge
Benefits of more data
Homomorphic Encryption
Homomorphic Encryption Progress
Homomorphic Classification Progress
Conclusion
Homework
Questions
Explanation Ability Scheme
Federated Learning
Adverse Setting
Taught by
RSA Conference
Related Courses
-
Federated Learning and Privacy-preserving RAGs
-
Privacy-Preserving Machine Learning
-
Privacy-Preserving Machine Learning: Strategies and Tools
-
Privacy Preserving Machine Learning with Fully Homomorphic Encryption
-
Privacy-Preserving Machine Learning with Fully Homomorphic Encryption
-
Preserving Data Privacy in Federated Learning - Xiaokui Xiao
4.0
