Claude Shannon and Perfect Secrecy: The Information Theory of Cryptography

Shannon turned secrecy into a question that mathematics could answer

Before Claude Shannon, cryptography already had centuries of practice, ingenious machines, statistical attacks and military doctrine. What it lacked was a general mathematical language for asking how much information a ciphertext reveals. Shannon supplied that language. His 1949 paper Communication Theory of Secrecy Systems, based on a classified 1945 report, treated secrecy systems using probability, entropy and the same information-theoretic framework he had developed for communication.^[1][5] The importance of the move was methodological: a cipher no longer had to be judged only by whether anyone had broken it. One could define a security property and prove whether a scheme satisfied it.

Security became a property of distributions

Instead of asking whether ciphertext “looks random,” Shannon asked whether observing it changes the probability assigned to possible plaintexts.

Perfect secrecy means the ciphertext teaches the attacker nothing about the message

Shannon’s definition of perfect secrecy can be stated as a requirement that the posterior probability of a message, after seeing the ciphertext, remain the same as the prior probability. In modern terms, the message and ciphertext are statistically independent.^[1] MIT’s current cryptography notes present the equivalent indistinguishability view: for any two candidate plaintexts, a ciphertext generated under a uniformly random key should be distributed identically.^[3] This is stronger than saying an attack is expensive. Perfect secrecy grants the adversary unlimited computing power and still demands that the ciphertext contain no usable information about which message was encrypted.

The one-time pad gave Shannon a concrete example of information-theoretic security

The one-time pad combines a message with a uniformly random key of the same length and never reuses that key. Under those conditions, every ciphertext is compatible with every possible plaintext through exactly one key. Shannon showed why this produces perfect secrecy; modern treatments use the same elementary argument.^[1][3] The system itself predates Shannon. NSA’s historical timeline credits Gilbert Vernam with a one-time tape teleprinter system in 1917, while later practice added truly random keying material.^[2] Shannon’s contribution was to prove exactly what security such a construction provides.

Perfect security depends on key discipline

If a one-time pad key is reused, ciphertexts can be combined to cancel the key and expose relationships among plaintexts. The mathematics assumes the operational rule is actually followed.

Shannon also proved that perfect secrecy has an unavoidable resource cost

The one-time pad is not merely one clever perfectly secret scheme. It illustrates a lower bound. In a perfectly secret encryption system, the key space must be at least as large as the message space under the usual correctness assumptions.^[3] That result explains why information-theoretic secrecy is difficult to deploy at scale. Encrypting gigabytes of unpredictable data with a one-time pad requires gigabytes of secret random key material that must already be distributed securely. The security proof does not remove the key-distribution problem; it exposes it as fundamental rather than an implementation inconvenience.

Information theory separated uncertainty from computational difficulty

Shannon’s 1948 communication theory formalized entropy as a measure related to uncertainty in a source.^[5] The secrecy paper then applied related ideas to cryptographic systems, considering plaintext statistics, keys, ciphertexts and equivocation.^[1] This created a conceptual distinction that remains central. Information-theoretic security asks what can be learned even with unlimited computation. Computational security instead assumes certain tasks require infeasible resources. Most modern cryptography relies on the latter because it allows short reusable keys and practical public-key protocols, but Shannon’s stronger benchmark remains the cleanest possible notion of secrecy.

Perfect secrecy is stronger than practical unbreakability

A computational cipher may be secure because the best known attack would take astronomical time. A perfectly secret cipher is secure even if the attacker has unlimited time.

The one-time pad’s strength makes its operational weaknesses unusually visible

NIST’s terminology still defines one-time cryptosystems around the requirement that keying material be used only once.^[4] That rule highlights the tension between mathematical and operational security. Keys must be generated with sufficient randomness, distributed secretly, protected at both endpoints, synchronized with messages and destroyed or retired after use. A perfect theorem does not protect a copied key, a predictable random generator or a reused pad. The one-time pad therefore became a classic example of a broader security lesson: a primitive can be mathematically ideal while the system around it remains difficult to operate safely.

Shannon gave cryptography a style of proof that later fields expanded

Modern cryptography routinely begins by defining an adversary, a security experiment and the information or advantage the adversary is allowed to obtain. Shannon’s perfect-secrecy definition is an early and unusually strong example of that style. Later work relaxed the requirement from exact statistical independence to computational indistinguishability, but retained the idea that security should be stated precisely enough to prove.^[3] This was a major change from treating a cipher as secure because its mechanism was obscure or because no successful attack had yet been published.

The definition can matter more than the algorithm

Once a field agrees on what “secure” means, algorithms can be compared against the same target instead of judged by reputation or complexity of design.

Why perfect secrecy belongs at the foundation of computer security

Shannon’s work belongs at the foundation of computer security because it turned secrecy into a formal relationship among messages, keys and observations. The 1949 paper connected cryptography with information theory, the one-time pad demonstrated that absolute secrecy is achievable under strict conditions, and the lower bound explained why that ideal is expensive.^[1][3] The result did not make practical cryptography simple. It made the tradeoffs explicit.

The continuing value of perfect secrecy is partly conceptual. It supplies a security notion with no hidden assumption about the attacker’s processor, algorithmic cleverness or future hardware. When modern cryptographers choose computational assumptions instead, they do so knowingly because perfect secrecy’s key requirements are often impractical.

That intellectual discipline is Shannon’s deeper legacy. Cryptographic trust should rest on a stated model and a reasoned argument about what an adversary can learn. The machinery may change—from tapes to block ciphers to post-quantum schemes—but the demand for a precise security claim begins with the same question Shannon made mathematical: what information does the ciphertext reveal?

RESEARCH / PROVENANCE

Works Cited

5 SOURCES

1. 01
   Claude E. Shannon — Communication Theory of Secrecy Systems ↗onlinelibrary.wiley.com
2. 02
   NSA — Cryptologic History Timeline ↗nsa.gov
3. 03
   MIT 6.5620 — Intro & Perfect Secrecy ↗mit6875.github.io
4. 04
   NIST CSRC Glossary — One-Time Pad ↗csrc.nist.gov
5. 05
   Claude E. Shannon — A Mathematical Theory of Communication ↗onlinelibrary.wiley.com

CodeHistory is a living archive. Citations document the evidence used for this edition; later evidence may refine the account.