SF-PCA: Scalable and privacy-preserving federated principal component analysis
Scalable and Privacy-Preserving Federated Principal Component Analysis
AI summary
SF-PCA is a secure system for federated principal component analysis over private data from multiple data providers, protecting the original data and intermediate results against passive attackers and in the case where at most all but one party collude. It combines multi-party homomorphic encryption, interactive protocols, and edge computing to interleave local plaintext and encrypted-data computation, achieving accuracy comparable to a non-secure centralized scheme and being 3x to 250x faster than schemes using only secure multi-party computation or homomorphic encryption.