← Samples

Unpredictable random numbers

12345678#include <random> int main() { std::random_device r; std::seed_seq seed_seq{r(), r(), r(), r(), r(), r()}; std::mt19937 engine{seed_seq}; }

This sample is licensed under the CC0 Public Domain Dedication.


Seed a random number engine with greater unpredictability.


A pseudo-random number engine is initialised with a seed, which determines the initial state of the engine. When seeding with a single 32 bit value (as is typically provided by std::random_device), the engine can be in one of only 232 states. By providing an an engine with a greater amount of random seed data, we increase the number of possible states and therefore the unpredicitability of the engine.

On line 4, we create a std::random_device as a source of non-deterministic random numbers (if such a source is available). On line 5, we initialise a std::seed_seq with several initial seeds produced by the std::random_device. On line 7, we seed a random number engine with this std::seed_seq, which provides the engine with a sequence of evenly distributed seeds with low bias (although not bias-free). Providing more initial seeds will increase the unpredictability of the engine by increasing the number of possible seed sequences that might be generated.

While a std::seed_seq can be used to initialize multiple random number engines, bear in mind that a particular std::seed_seq will produce the same sequence of seeds each time it is used.

Note: An unpredictable random number engine is not necessarily cryptographically secure.


  • Joseph Mansfield
  • Seth Cantrell

Last Updated

18 April 2015


Fork this sample on GitHub