0

I thought I had found an answer to my question when I read this:

link

There, one of the posters effectively claims that your seed word's gematria mod 10 is a random number. In other words, sum the letters of a word, treating each letter as the number of its place in the alphabet. Then divide it by 10 and the remainder is your random number.

Another poster even "proves" this digit is random by showing a graph representing a flat distribution of digits after inputting thousands of common English words as seeds.

But when I tried this, I got a very UNFLAT distribution. 9 and 1 are TWICE as common as any other number except 3.

So, is there a practical way to manually (with pen and paper) generate a cryptographically-secure pseudo-random number using natural text as a seed?

Girkov Arpa
  • 1
  • 1
    Cryptographically secure RNGs must, by definition, have some physical source of real entropy. Modern CPUs have special hardware for it; before that, OSs would use things like keyboard and mouse timing. Whatever procedure you use *must* involve flipping coins, rolling dice, or something similar, otherwise it's just a hash function of the input text. – Lee Daniel Crocker Aug 25 '18 at 21:00
  • I would be happy with a practical pen-and-paper hash function with which to generate a key suitable for a one-time pad. Am I asking for something impossible? – Girkov Arpa Aug 25 '18 at 21:16
  • If you insist on using natural text as seed, it is impossible. Even if you get something looking random, it cannot be cryptographically secure, because a deterministic algorithm will always reflect the structure of the input to the output. – Adrian W Aug 25 '18 at 21:29
  • Wouldn't an attacker have to first figure out what algorithm I used for that principle to even matter though? I wrote a script that converts natural text into a simple string of lowercase letters in groups of 5. The gematrias of each group (the sum of the positions of their letters in the alphabet) modulus 10 are concatenated into a string of digits. The frequency of each digit is 10%. If I use this for a one-time pad how on earth would anyone be able to crack it? – Girkov Arpa Aug 25 '18 at 22:50

0 Answers0