One of the biggest flaws in enciphering text with a
transposition cipher is that all of the letters of the
plaintext appear in the ciphertext, with the exact same
This idea is a method of adding padding to the
plaintext, such that the letter frequencies of the
not known to the attacker.
Step zero: Before sending any plaintexts, the person
who will be doing the encryption should do a simple
analysis of a corpus of text in the same language as the
messages will be in, and determine the relative
frequencies of the symbols. This set of relative
frequencies is called the frequency model. This step is
done once ever, not before every message.
Step one: Alter the plaintext in whatever manner is
necessary so that the proper recipient can determine
where it ends.
Step two: Determine the relative frequencies of the
symbols in the plaintext.
Step three: Identify those symbols which the plaintext
is deficient in, compared to the frequency model.
Step four: Append to the plaintext those symbols that
the plaintext doesn't have enough of, until the relative
frequencies of symbols of the plaintext is sufficiently
close to the frequency model.
Step five: Encipher using some form of transposition
Because all plaintext messages were padded to closely
match the frequency model, all ciphertext messages will
also match the frequency model. Thus, all ciphertexts
will look very much like one another.