Unbreakable Code?

by Ich

A fun code is the box code. Can be heavily tweaked for maximum security.

Table 1:  Alphabetical


a b c d e a
f g h i j f
k l m n o k
p q r s t p
u v x y z u
a b c d e a

Let's encrypt "BTW" for lack of anything better, and because the letters illustrate how some aspects of the code are used.

Find the "B" that is furthest up and left. If you draw a 2x2 box around it and its neighbors that are down and right of it, you have the letters c g h (omitting B, naturally). Now, pick any two of them. cg, ch, gh, gc, hc, hg. Six ways to write one letter. T and W could be represented as follows:

pb, pu, bu, bp, up, ub and xb, xc, bc, bx, cx, cb

Nifty? We're just getting started.

Find them on the chart, put a 2x2 box around them, and take the letter in the upper left corner.

There are 216 different representations of the plaintext "BTW" (6^3)

When you write into box, remember to NEVER use the same spacing patterns as you did in the original message, and NEVER use punctuation. These are IMPORTANT in all ciphers, but in box the first has added importance. For one, people will notice that all your words are multiples of two in length, and some are very long. To be deceptive, try putting them all into 4 letter groups:

CGAB TZIN NOYT FGIJ TPRM ZUIK becomes "Bush is a dolt" But, you say, there's two too many characters to be that. Dummy characters are useful to put into any cipher at the end to pad it out and throw people's efforts. XD VSK FFNSP XNV is garbage (I think), and can't be translated into English because there's nothing to translate. And if people measure your frequency and structure while it's combined with this garbage, it messes everything up. To further screw up people, our original "Bush is a dolt" can become "CG ABTZI N NOY'T FG IJT PRM ZUIK." When you decipher it, just make it "bushisadolt" and don't worry about spaces. If you're really concerned about the message changing, insert an x or something in between the words.

Notice, in the above chart, I used v for w. You can include your 25 favorite letters of the alphabet in the inner 5x5 box. The outside is just tacked on so you can do the edges.

This code expands your message in length by 2x, and also gives you consistently even message lengths (counting by twos, of course) but these indicators about your code aren't really important and can be overcome. This can get time consuming, but if you wanted *easy* then you'd pick a fucking substitution cipher.

A large benefit about this is you can't just "frequency count" the characters. If someone was astute enough to frequency the two-character-sets, they'd get 150 results in a message that uses each representation of your 25 favorite letters at least once. This makes this code *far* harder to crack.

Going further with this:

Numbers: If you're not doing something extremely technical, like listing IP addresses, you could write NUM ABC, where A=1, B=2... and ABC is 123, your number. You can create your variation, or make NUM be represented a pair of two specific characters, "az" for example.

Even further protection: Merging

Cut the code in halves.
CGABTZINNOYTFGIJTPRMZUIK becomes
CGABTZINNOYT FGIJTPRMZUIK

Place the halves atop one another, and merge

C G A B T Z I N N O Y T 
 \/\/\/\/\/\/\/\/\/\/\/\
 F G I J T P R M Z U I K

Voila! CFGGAIBJTTZPIRNMNZOUYITK!

For added security and more time spent enciphering and deciphering the code: you can also arrange ahead of time with your partner in crime to use a defined box that looks something like this:

Table 2:  Scrambled


a n s h d a 
p k e b t p
y i o g v y
q c l r z q
x m j p f x
a n s h d a

BTW could be written G TV YZY

You can also throw in a pattern of "use the UL corner, use UR, LL, LR, UL, LL etc.", where UR means b

Example: If you're encoding the letter "O" using the second chart, the first representations would be: lg, lr, gr, gl, rl, rg; the second: il, ic, lc, li, ci, cl.

You can put in pages of "junk" text and tell someone to start translating where the see the letters "tz" together, end at zt. Like a tag in html, everything between tz and zt is affected. Search through your document to avoid accidentally having an unintended "tz". Your messagee could search, manually or with computer, through the document to find the open tag. 6000 characters of crap and 60 of substance will all but thwart the efforts of someone who isn't on the inside. At about this point your greatest threat is from someone tapping your phone or sneaking around your house/computer while you're out, rather than break into your code, and deciphering is getting to be such a pain that MD5 is looking pretty appealing. If you use a dynamic table that changes (all letters advance 1 in the alphabet etc.) then you've added another degree of security.

The important reminders are to NEVER retain spacing or punctuation, always try to use different combinations of the letters, use a "scrambled box code" whenever possible, use dummy characters (two-sets of letters that don't fall in a 2x2 box, like "pc" above, or even two from a 2x2 box but where one of them is the "read" character, like at a place where you'd be reading the Upper Left letter and have the letter set "bv" and you've got a dummy character, or you could use "bb".

The brilliance of a code isn't what protects it from being cracked. Idiots using it make it cracked. Germans in WW2 broadcast the message "a bird in the hand in worth two in the bush" [encoded in German] or some such saying rather than "Nothing to report"

And lastly, other precautions should be taken as well. If you let your ciphered message fall into the enemy's hands, you create--rather than increase--the possibility that it will be decoded.

If you didn't understand part of this, check the following:

1. You are not an idiot.
2. You started reading at the beginning, and understand that I wrote this at a time when I should have been sleeping, but wasn't.

Questions can be directed to my email address, ibd[x]n@yahoo.com x=wag

I'll try to answer them.