Generate super random number in c++

Question

C++11 Introduced the class that allows for generating very random numbers, it also creates an even distribution of random numbers. There is also implementation to generate a seed (a number used to make the Random Number Generator more random).

I am trying to make a function that generates a random number between min and max but I am having trouble. The function only generates the seeds and the random number once. When I call the function in other words it will keep giving me the same number.

Below is the code, I try to generate a bunch of seeds, pick one of them randomly, use that seed for the RNG and finaly produce a random number.

int Utils::GenerateSuperRandomNum(int min, int max)
{
    //Seed a the RNG
    int randNum;
    int randIndex;
    seed_seq seq{ 1, 2, 3, 4, 5 };
    vector<int> seeds(5 * max);
    uniform_int_distribution<int> rngDistribution(min, max);    //Generates number in the range min to max.

    //Generate our seed numbers.
    seq.generate(seeds.begin(), seeds.end());

    //Generate random index bewteen 0 and size - 1.
    srand(seeds.at(0));
    randIndex = rand() % seeds.size(); 

    //Seed the RNG with a random seed from our vector.
    mt19937 rngGenerator(seeds.at(randIndex)); 

    //Get a random number.
    randNum = rngDistribution(rngGenerator);

    return randNum;
}

Answer 1

seed_seq seq{ 1, 2, 3, 4, 5 };
vector<int> seeds(5 * max);
uniform_int_distribution<int> rngDistribution(min, max);    //Generates number in the range min to max.

//Generate our seed numbers.
seq.generate(seeds.begin(), seeds.end());

seq is always fed the same input {1,2,3,4,5} so always has the same state. Since it has the same state, seeds.at(0) is always the same value.

//Generate random index bewteen 0 and size - 1.
srand(seeds.at(0));

Since srand is seeded with that same value every time, it to starts with the same state every time. Since it receives the same value every time, it always starts with the same state. The same state that rand uses.

randIndex = rand() % seeds.size(); 

Since rand always has the same state as per the srand, it will always generate the same first number every single time.

 mt19937 rngGenerator(seeds.at(randIndex)); 

Since randIndex is always the same value, then seeds.at(randIndex) is always the same value. Since rngGenerator is always seeded with the same value, it always has the same state.

randNum = rngDistribution(rngGenerator);

Since rngDistribution always has the same state, it always produces the same value.

This is obviously a problem. The simple fix is to seed based on the CPU temperature, or the time, or some other value that changes often.

---

Basically, you've seriously overthought this. It's designed to be used like this:

int Utils::GenerateSuperRandomNum(int min, int max) {
    static mt19937 rngGenerator(std::random_device{}());
    std::uniform_int_distribution<int> rngDistribution(min, max);
    return rngDistribution(rngGenerator);
}

std::random_device{}() generates a vaguely randomish number based on magic, hopefully hardware, like maybe the CPU tempurature or something. It can be slow or have other issues, so you should only use it once per program. Namely, to seed a faster/better generator.

static mt19937 rngGenerator( creates a single global generator, which is seeded the first time the function is called, and is never seeded again. This is perfect, because we want it to be initialized once, and then just do it's magic from then on. The usage of other generators like rand don't add any entropy, so don't do that. We also don't want to re-seed it, as that might accidentally reduce randomness, rather than increase it.

std::uniform_int_distribution<int> rngDistribution(min, max); and rngDistribution(rngGenerator) you seem to understand. They use the generator to give random numbers in that distribution. Done deal.

Answer 2

A seed does not make a (pseudo-)random number generator more random. It provides a starting point for generating a reproducible sequence of random numbers.

Which means, if you provide the exact same seed, you'll get the exact same results.

Answer 3

Making a random number generator more "random" by using it to generate it's own seed is a bit like trying to lift yourself up by pulling on your boot straps. If this is just for fun the time is a sufficiently random seed, if you are doing super spy crypto a h/w device that provides true random events is what you need.

You could try using this service, I don't know if it is valid or just the NSA trying to trick you. https://www.random.org