Safedisc

[caki]

While I am not a crypto guy, and have almost no knowledge of how crypto works, wouldn't it be possible, in the age of modern computing, to make an ubreakable code with something based off of a one-time pad?

Consider this:

Lets say we want to protect a text with letters A-Z and " " in it. Okay, thats 27 characters we need to worry about. Now generate all the numbers in between 1 and 27000. To each of the characters we want to protect, assign a 1000 random numbers from the collection of numbers in between 1 and 27000, so:

A = 47, 51, 26999...
B = 21, 407, 1096...
etc

Write all of this into a file, and this is your key. Now, when you encrypt a string, you take the character you are encrypting, for example "A", and then replace it with one of the 1000 random numbers assigned to it.

Heres an example:

String to encrypt:
HELLO WORLD

H - 34, 59...
E - 99, 127...
L - 89, 56, 949...
O - 23, 77...
W - 909, 230...
R - 654, 767...
D - 378, 3...
" " - 21, 90...

So encrypted one time, it could be:
34 99 89 949 77 21 909 23 654 56 3

Encrypted another time, it could be:
59 127 56 949 23 90 230 23 767 56 3

Its different every time, but decrypt to the same thing with the correct key, there are a 1000 numbers to each letter, so even if there are duplicates they would be hard to correlate... it would be pretty hard to break small messages encrypted this way, providing the key is never found... for larger messages just increase the key size. If you keep the key size/encrypted file size at a certain ratio, change the keys often enough and/or use large enough keys, and have a good random number generator, wouldn't this be pretty secure?

These are just random musings that I had during math class a year ago, but I'm interested in seeing what other people think.

[Joe Forster/STA]

This is only another implementation of the Caesar cipher which is one of the earliest and most simplicistic ones. It can be deciphered easily for long encrypted messages (i.e. text files, sequences of letters), comparing the frequency of each numerical code to the frequency of each letter in the target alphabet (e.g. in English the most frequent letter is E, then come, I think, T and S etc.) to find out which code stands for which letter. But you're right, many codes for the same letter and short messages make the decryption harder. However, there are always geniuses who can crack such messages in surprisingly short times, with intuition, tricks, heuristics, trial and error and lots of calculations.

Actually, the Germans used a similar approach in the second world war but they assigned codes to words rather than letters. (The frequency is words is pretty much impossible to calculate...) Of course, the allies managed to find out lots of words but many others were still missing. At the end, they found a complete code book somewhere (on a U-boat?) that the Germans forgot to destroy. (There was an interesting documentary about this on National Geographic or Discovery Channel, try to find it.) Was this the famous Enigma?!