Cryptography Reference
In-Depth Information
Networked Computers
The advent of networked computers signaled a remarkable expansion of
the cryptographic field, as the financial and governmental sectors began
to assess their needs for secure electronic transactions. In the early 1970s,
IBM researcher Horst Feistel—at the time one of the few private-sector
cryptographic researchers—began work on an algorithm that would, he
argued, safeguard individual privacy in the coming computer society.
Feistel proposed that ciphers could be used to encrypt databanks in order
to guard their contents from anyone but authorized individuals. Feistel
presented these ideas and the design principles of his Lucifer cipher in a
1973
Scientific American
article, at the time one of the most explicit discus-
sion of modern cryptographic principles ever presented to the American
public.
47
The design of Lucifer was remarkably simple: like the Vernam cipher, it
worked directly with the binary representation of messages: “We shall
tacitly assume that the messages we wish to handle cryptographically have
first been translated into a sequence of binary digits. Any kind of informa-
tion, be it letters, musical sounds, or television signals, can be represented
by binary coding. We shall no longer worry about the original meaning of
the messages. We shall deal only with their binary representation.”
48
It relied on Shannon's suggestion in
Communication Theory of Secrecy
Systems
that strong and practical ciphers could be constructed from two
simple methods—
diffusion
and
confusion—
which, when combined, made
difficult any statistical (i.e., frequency) analysis of the ciphertext.
Diffusion dissipates the redundancy found in natural languages by
spreading it throughout the ciphertext.
49
A simple permutation of the
letters of the plaintext, for example “hello” to “lohel,” breaks the frequen-
cies of bigrams and trigrams. Confusion makes the relation between the
key and ciphertext as involved as possible, so that statistical analysis of the
cipher text will not yield much information about the key. Mono- and
polyalphabetic substitution are both methods of achieving confusion. Per-
mutation and substitution are inadequate methods of encryption by them-
selves, but Shannon showed that their combination yields practical ciphers
of great speed and resilience. The Lucifer ciphering process consisted pre-
cisely in the interleaving of successive operations of diffusion and confu-
sion. The algorithm performs sixteen rounds of such operations on the
