3 ms·
With Classic McElice, at least right now, the computation time for a keypair is a bigger problem than the size of the public key. It will get faster with better
by drummer 6y ago
With Classic McElice, at least right now, the computation time for a keypair is a bigger problem than the size of the public key. It will get faster with better implementations and hardware use. But now it takes many seconds on a Ryzen 9 3950 to generate a keypair. I had to build a system with background threads computing keys and having them ready for quick response when a connection is made.
If anyone is interested see: https://github.com/kareldonk/QuantumGate https://github.com/kareldonk/QuantumGate
- dependenttypes 6y agoYou might want to check out https://groups.google.com/a/list.nist.gov/forum/#!topic/pqc-forum/koZWqTM714A https://groups.google.com/a/list.nist.gov/forum/#!topic/pqc-... This project that you linked seems quite nice. I especically like its Code of Conduct.
- drummer 6y agoThanks for the link. It seems the implementations are already improving. From the posted benchmarks it looks like it got twice as fast, which is still slow for some use cases (like ephemeral use for forward secrecy) and slower devices. The best solution will be hardware support.