Re:Fully Homomorphic Encryption Without Bootstrapping [pdf]
by deegles on Wednesday June 17, 2015 @02:32PM
(#9734242)
How many bit operations do non-FHE tasks need in comparison?
Re:Fully Homomorphic Encryption Without Bootstrapping [pdf]
by themeek on Wednesday June 17, 2015 @02:38PM
(#9734287)
Admittedly much fewer (on the order of 'a few'). It's also hard to compare, as constants are hidden in big-O. Real performance comparisons between cryptosystems are comparisons of real-world engineering implementations: especially since side channel blinding, exception and case handling, etc need to be taken into effect.
The take away is that this would still be nowhere near as efficient as 'traditional', especially symmetric, encryption. What would be a breakthrough is that presumably the scheme would be orders of magnitude more efficient than current existing theoretical FHE schemes.
Evaluating the claim thus requires skepticism and care. The quality of the paper is suspect as are its proofs. But, as is the case with science, heuristics like this count for little.
They do seem to have selected a hard worst-case problem to base the system on. Obviously this means very little if secure keys and setting in this complexity space can not be found in practice - or if the details of the cryptosystem, for one of many many reasons, lead to its easy compromise.
Looking forward to a proper peer review of the scheme.
Edit: Looks like it's already broken!
https://news.ycombinator.com/item?id=9734512