Communication Theory of Secrecy Systems?
By C. E. SHANNON
1 INTRODUCTION AND SUMMARY
The problems of cryptography and secrecy systems furnish an interesting application of communication theory1. In this paper a theory of secrecy systems is developed. The approach is on a theoretical level and is intended to complement the treatment found in standard works on cryptography2. There, a detailed study is made of the many standard types of codes and ciphers, and of the ways of breaking them. We will be more concerned with the general mathematical structure and properties of secrecy systems.
The treatment is limited in certain ways. First, there are three general types of secrecy system: (1) concealment systems, including such methods as invisible ink, concealing a message in an innocent text, or in a fake covering cryptogram, or other methods in which the existence of the message is concealed from the enemy; (2) privacy systems, for example speech inversion, in which special equipment is required to recover the message; (3) “true” secrecy systems where the meaning of the message is concealed by cipher, code, etc., although its existence is not hidden, and the enemy is assumed to have any special equipment necessary to intercept and record the transmitted signal. We consider only the third type—concealment system are primarily a psychological problem and privacy systems a technological one.
Secondly, the treatment is limited to the case of discrete information where the message to be enciphered consists of a sequence of discrete symbols, each chosen from a finite set. These symbols may be letters in a language, words of a language, amplitude levels of a “quantized” speech or video signal, etc., but the main emphasis and thinking has been concerned with the case of letters.
The paper is divided into three parts. The main results will now be briefly summarized. The first part deals with the basic mathematical structure of secrecy systems. As in communication theory a language is considered to be represented by a stochastic process which produces a discrete sequence of ? The material in this paper appeared in a confidential report “A Mathematical Theory of Cryptography” dated Sept.1, 1946, which has now been declassified.
1 Shannon, C. E., “A Mathematical Theory of Communication,” Bell System Technical Journal, July 1948, p.623.
2 See, for example, H. F. Gaines, “Elementary Cryptanalysis,” or M. Givierge, “Cours de Cryptographie.” symbols in accordance with some system of probabilities. Associated with a language there is a certain parameter D which we call the redundancy of the language. D measures, in a sense, how much a text in the language can be reduced in length without losing any information. As a simple example, since u always follows q in English words, the u may be omitted without loss. Considerable reductions are possible in English due to the statistical structure of the language, the high frequencies of certain letters or words, etc. Redundancy is of central importance in the study of secrecy systems.
A secrecy system is defined abstractly as a set of transformations of one space (the set of possible messages) into a second space (the set of possible cryptograms). Each particular transformation of the set corresponds to enciphering with a particular key. The transformations are supposed reversible (non-singular) so that unique deciphering is possible when the key is known.
Each key and therefore each transformation is assumed to have an a priori probability associated with it—the probability of choosing that key. Similarly each possible message is assumed to have an associated a priori probability, determined by the underlying stochastic process. These probabilities for the various keys and messages are actually the enemy cryptanalyst’s a priori probabilities for the choices in question, and represent his a priori knowledge of the situation....