Seems like still 7-orders of magnitude slower than native. Thats is progress though and 1-minute for a single AES block might start to have some niche areas of use if there are no direct algorithms to do whatever it is that needs to be done. (Plus a bunch of esoteric crypto stuff and hardness assumptions that might get weakened over time.) Adam On Sat, Jan 18, 2014 at 02:01:56AM -0800, coderman wrote:
http://eprint.iacr.org/2014/039.pdf [see pdf for citations / bib]
Gentry, Halevi and Smart introduced the first evaluation of a complex circuit, i.e. a full AES block evaluation [...] With 5 minutes per block evaluation time the byte-sliced implementation is faster, [...] homomorphically evaluate the full 128-bit AES circuit in a bit-sliced implementation to demonstrate the scalability of the introduced technique. Our implementation is 5 times faster than the byte sliced implementation