-1
$\begingroup$

Say I want to seed a CSPRNG. What sources are popular, but unreliable? What sources should I even use?

$\endgroup$
  • $\begingroup$ it will be good to combine this with your previous question. $\endgroup$ – hardyrama Sep 1 '19 at 3:16
  • 3
    $\begingroup$ Any language-specific rand() is commonly (mis)used as a cryptographic seed. $\endgroup$ – forest Sep 1 '19 at 7:30
  • 3
    $\begingroup$ Using the current time and date is popular, but unreliable. Oh, abd also totally insecure. $\endgroup$ – Frank Denis Sep 1 '19 at 16:20
0
$\begingroup$

If it is for your personal use, randomness collected from mouse movements can be an option. Veracrypt uses this. If it is to be done on a large scale, try using Lava Lamps. If these seem unviable, try using the Urandom pool.

|improve this answer|||||
$\endgroup$
  • $\begingroup$ Do you have "What sources are popular, but unreliable?" the right way around? Lava lamps are a pretty perfect source of entropy (Rayleigh–Bénard convection/deterministic chaotic systems). I also suggest changing to /dev/random if you're looking for perfect but unreliable entropy sources. $\endgroup$ – Paul Uszak Sep 1 '19 at 23:39
  • $\begingroup$ It's important to realise that Thomas Hühn's silly 'myth' post is just a flawed and dangerous internet mime that seems to catch out young players. See and digest Mr. Pornin's acute comment in said post regarding the fundamental difference that is appropriate to this (and all other) question(s). $\endgroup$ – Paul Uszak Sep 2 '19 at 0:16
  • 2
    $\begingroup$ @PaulUszak That "myth" post is not flawed. As someone who has read both the current LinuxRNG documentation and is familiar with the driver source code itself, I can say the only issues his post has is a few oversimplifications and a tiny mistake caused by him missing a change to the driver that came in kernel 4.8. But the idea behind is post is absolutely correct. Your view that you need the blocking pool to be secure is simply wrong. It's only perfect in theory if the entropy estimation is perfect, which it is not. $\endgroup$ – forest Sep 2 '19 at 4:20

Not the answer you're looking for? Browse other questions tagged or ask your own question.