How do we measure and compare the speeds of NTRUEncrypt, RSA and ECC?
Statements like Cryptosystem XXX is 100 times faster than Cryptosystem YYY are meaningless without additional information. Unfortunately, the marketplace insists on succinct quantitative comparisons, so we and our competitors frequently publish these sorts of statements. To be accurate, three issues need to be addressed: Implementation The operating speed of a cryptosystem depends heavily on the particular implementation, on the compiler, on the machine or device processor, and on a host of other similar factors. To cite an illuminating example, we have found that the excellent weidai crypto++ package runs half as fast on our Pentium machines as it does on the weidai Celeron machines. Optimizations for a particular machine or using various programming tricks may easily give speed differentials of 2 to 3 times. We also note that we are comparing fully optimized implantations of RSA and ECC with the implementation of NTRUEncrypt in the first commercial release of Tumbler. We certainly ex
