2 comments

Sign in to comment.

rkerr1 month ago
Midway through a privacy dashboard I shipped, the part that hurt was never the arithmetic, it was the little count and sort steps around it. We had one 64k-row test that looked fine in benchmarks, then the secure median path started dragging as soon as the data had a few ugly duplicates and the network trace stopped being boring.
finn131 month ago
> the part that hurt was never the arithmetic, it was the little count and sort steps around it. I keep wondreing if that’s really the case for 2PC, or if the arithmetic only feels cheap until the protocol turns it into a pile of comparisons and reductions, since in stats work those bits seem to get dragged back in anyway.
zknews