Explore a 16-minute conference talk from USENIX Security '23 that introduces ACORN, an innovative protocol for secure aggregation with input validation. Learn how this new approach improves computational efficiency and extends secure aggregation capabilities. Discover how ACORN enables servers to verify client input constraints, such as L0, L2, and Linfinity bounds, preventing malicious clients from exerting undue influence on aggregate statistics or machine learning models. Examine the protocol's performance improvements, including 2-8X speedups in client computation and over 30X reduction in client communication compared to prior work. Gain insights into ACORN's practical applications, such as processing binary indicator vectors and dense vectors efficiently. Understand how this advancement contributes to privacy-preserving distributed statistical analysis and machine learning.
Overview
Syllabus
USENIX Security '23 - ACORN: Input Validation for Secure Aggregation
Taught by
USENIX
Related Courses
-
ELSA - Secure Aggregation for Federated Learning with Malicious Actors
-
Flamingo - Multi-Round Single-Server Secure Aggregation with Applications to Private Federated Learning
-
Duoram - A Bandwidth-Efficient Distributed ORAM for 2- and 3-Party Computation
-
Squirrel - A Scalable Secure Two-Party Computation Framework for Training Gradient Boosting Decision Tree
-
GAZELLE - A Low Latency Framework for Secure Neural Network Inference
-
SecretFlow-SPU - A Performant and User-Friendly Framework for Privacy-Preserving Machine Learning
