Secure multi-party linear regression at plaintext speed
Jonathan M. Bloom, Hail Team / Neale Lab / Stanley Center for Psychiatric Research, Broad Institute of MIT and Harvard
We detail distributed algorithms for scalable, secure multi-party linear regression and feature selection at essentially the same speed as plaintext regression. While the core geometric ideas are simple, their utility in combination for feature selection appears to be novel. Our scheme opens the door to efficient and secure genome-wide association studies across multiple biobanks.
We recently found that the approach to SMC linear regression in Section 2 has been explored since at least 2005, c.f. Secure Regression on Distributed Databases.
Run a Python demo of the multi-party linear regression algorithm in Section 2.
Hail uses the single-party distributed algorithm in Section 3 to enable massive genomic analyses and will include multi-party algorithms someday. Also check out the exciting work on secure genomics by Hoon Cho, Bonnie Berger and colleagues.
Feedback welcome! Write Jon: firstname.lastname@example.org