Cryptography Reference
In-Depth Information
TUIMERMZXPVEWFYOZHYVMIDLHVECI
- ANEXTREMELYLONGANDCOMPLETELYR
= THEPLAINTEXTISSOMEWHATSHORTER
We encrypt bitwise rather than bytewise nowadays. We no longer add characters
modulo 25, but bits modulo 2, which corresponds to the bitwise exclusive OR
(XOR, or operator in the C programming language). That's a basic operation for
every microprocessor, and it can also be used to encrypt arbitrary data streams.
The important point about this method is that the section of the key concerned
may be used once only. Since absolutely nothing is known about the key, any
plaintext could have produced a given ciphertext, and all conceivable plaintexts
are equally likely. All other symmetric methods (these are encryption algo-
rithms like the ones described in this topic so far) have shorter keys, which
means that they necessarily contain some rules (the trouble is, we just aren't
clever enough to recognize and exploit these rules). There are no such rules
with the one-time pad, not even in theory.
Unfortunately, there are two problems inherent in this method.
Problem 1
: How can we create a 'truly random' key? Using a computer won't
do the trick properly, because every computer output obeys rules, which means
that it can be at best 'pseudo-random'. What we need is chaotic input from
the real world, which cannot be traced back. How about this for 'chaotic' and
'real world': record whatever an unreliable Geiger counter measures from a
radioactive sample in the trunk of your car as you rumble over bumpy roads,
then overlay this data stream with the digitized gurgling of a waterfall and the
bleating of sheep. Every spy will just give up.
Since there were times when digitizing didn't exist, people selected sections
from topics for use as keys, for example. Naturally, that's far from being secure.
[BauerDS] explains the so-called
zigzag method
, which
does
break this method
after all: for a starter, you know a little bit of plaintext, from which you can
easily calculate a fragment of the key. This fragment doesn't normally end at
word boundaries so that, with some luck, fragments missing in words can be
filled in. This, in turn, produces a little piece of plaintext — and so on. Though
this cryptanalysis doesn't always work, it can indeed reveal important parts of
the plaintext.
Good keys, i.e., non-reproducible keys, were carried around by good spies in
printed form. Adversaries would regularly pull them out of the hollow knob of
their walking sticks.
