"The higher the entropy of a random source, the better the quality of the random data it generates."

Many cryptographic algorithms call for a random source, either in key-generation, or some other primitive. Implementors must be extremely cautious in selecting that random source, or they will open themselves up to attack. For example, the only formally proven encryption technique, the one time pad, requires a completely random and unbiased key-stream that is at least as long as the message itself, and is never reused. There are many implicit complications presented in this requirement, as the only sources of "true randomness" are in the physical world (silicon decay is an example), and are impossible to implement in software. Thus, it is often only feasible to obtain pseudo-randomness. Pseudo-Random Number Generators, or PRNGs, use multiple sources that are thought to be difficult to predict (mouse movement, least significant digits of the computer clock, network statistics, etc.) in order to generate an entropy pool, which is passed through assorted algorithms which attempt to remove any biases, and then used as a seed for a pre-determined static set of numbers. Even with all of the sources of entropy, a determined attacker can usually reduce the effective strength of an implementation by cutting out some of the factorsâ€”for instance making educated guesses on the time. PRNGs that are thought to be acceptable for cryptographic purposes are called Cryptographically-Secure Pseudo-Random Number Generators, or CSPRNGs.

## EntropyEdit

In terms of information theory, entropy is defined as the measure of the amount of information expressed in a string of bits. For example gender contains 1-bit of entropy as it can be represented using a 1 for males and a 0 for females. The quality of a random source is determined by just how much entropy it generates, if the entropy is less than the actual number of bits then there is some repetition of information. The more information that is repeated, or the shorter the period of some PRNG, the lower the entropy and the weaker and more predictable the source of randomness. Therefore in cryptography one seeks to get as close to perfect randomness as possible with the resources available - where a perfect random number generator creates a sequence of bits which are unpredictable no matter how large a sample of previously generated bits is obtained.

## Further readingEdit

- Cryptography/Random number generation
- RFC 4086 "Randomness Requirements for Security"