• bitfucker@programming.dev
    link
    fedilink
    arrow-up
    2
    ·
    8 days ago

    Thanks! I was curious because I am using keepass and I do use argon2id lol. I was a bit skeptical because the 2s benchmark parameter seems oddly small, but nice knowing that it is actually quite good.

    • mlg@lemmy.world
      link
      fedilink
      English
      arrow-up
      2
      ·
      6 days ago

      On a GTX 1660:

      -------------------------------------------------------------------------------
      * Hash-Mode 34300 (KeePass Argon2 (KDBX v4)) [m=65536, t=2, p=2] [Kernel: pure]
      -------------------------------------------------------------------------------
      
      Speed.#01........:      434 H/s (75.99ms) @ Accel:67 Loops:4 Thr:32 Vec:1
      
      -------------------------------------------------------------------------------------------------------------
      * Hash-Mode 70000 (Argon2id [Bridged: reference implementation + tunings]) [m=65536, t=3, p=1] [Kernel: pure]
      -------------------------------------------------------------------------------------------------------------
      
      Speed.#*.........:       39 H/s
      
      --------------------------------------------------------------------
      * Hash-Mode 36100 (yescrypt) [N=4096, r=32, p=1, t=0] [Kernel: pure]
      --------------------------------------------------------------------
      
      Speed.#01........:      431 H/s (164.40ms) @ Accel:264 Loops:2048 Thr:32 Vec:1
      

      For some context, my keepassxc on my PC registers about t=50 for a 1 second unlock delay, so the example benchmark t=2 is faster than an IRL example, hence why the reference argon2id is much lower with only an additional transform round (t=3).

      But yeah it’s definitely good, and since argon2id allows you to set the parameters (memory, transform rounds, parallell) you can go make it hilariously indestructible if you feel like waiting like 30 seconds to unlock a DB lol.