Beaver Triples enable multiplication of secret-shared values in MPC without increasing the reconstruction threshold, enabling private multi-party computation.
Key Takeaways
Secret sharing is linear; addition of shares is free, but multiplying shares naively doubles the polynomial degree and raises the reconstruction threshold.
Beaver Triples [a],[b],[c] where c=ab act as one-time random masks; revealing d=x-a and e=y-b exposes nothing about x or y.
The identity xy = c + bd + ae + de lets parties compute [xy] from opened scalars d,e and existing shares, keeping degree constant.
Triples must be fresh per multiplication; reuse leaks relationships between secret inputs via d’-d = x’-x.
Stoffel’s MPC platform uses precomputed Beaver Triples (sourced from satellite entropy) to ship privacy-preserving compute features without exposing raw user data.
Hacker News Comment Review
No substantive HN discussion yet; the only comment flagged a severe mobile performance issue causing Safari to crash on iPhone 16, suggesting frontend rendering problems on the Stoffel site.
