10gic / modular-exponentiation-benchmark

Introduction

Modular exponentiation is a time-consuming operation, and this project has performed a Golang benchmark test on three different implementations.

Method	Note	Is Constant-Time	Run Time (Small Parameter)	Run Time (Big Parameter)
ModExp	Use Golang builtin big package	No 😢	17532 ns 😊	2281780 ns
ModExpGo3rdParty	Use cronokirby/saferith package	Yes 😊	28941 ns 😢	6000609 ns 😢
ModExpOpenSSL	Use OpenSSL (BN_mod_exp_mont_consttime)	Yes 😊	20389 ns	1325081 ns 😊

1. Constant-time is important for security because it prevents attackers from using timing analysis to deduce sensitive information about the input data
2. Test CPU: Intel(R) Core(TM) i7-4770HQ CPU @ 2.20GHz
3. Small Parameter: 256 bits base/96 bits exponent/256 bits modulus
4. Big Parameter: 2048 bits base/768 bits exponent/2048 bits modulus

Conclusion

1. If you are using it in a cryptography scenario, please choose ModExpGo3rdParty or ModExpOpenSSL.
2. If you are using it in a cryptography scenario and performance is a major concern, please choose ModExpOpenSSL.

Test

$ go test -bench .
goos: darwin
goarch: amd64
pkg: github.com/10gic/modular-exponentiation-benchmark
cpu: Intel(R) Core(TM) i7-4770HQ CPU @ 2.20GHz
BenchmarkModExp256-8                       69508             17532 ns/op
BenchmarkModExpGo3rdParty256-8             41390             28941 ns/op
BenchmarkModExpOpenSSL256-8                58264             20389 ns/op
BenchmarkModExp2048-8                        520           2281780 ns/op
BenchmarkModExpGo3rdParty2048-8              199           6000609 ns/op
BenchmarkModExpOpenSSL2048-8                 904           1325081 ns/op
PASS
ok      github.com/10gic/modular-exponentiation-benchmark       9.169s