Math.random vs other random algorithms

Question

I'm using a node.js client application to fuzz a remote server.

I can ~100% of the time crash the remote server when using Math.random in the node.js client. However, I've since tried a couple of deterministic, seeded random number generators, and neither are able to crash it.

I suspect its due to an idiosyncrasy of Math.random or the seeded generators I've tested.

This is one of the seeded generators I've tried:

var x = 123456789, y = 362436069, z = 521288629, w = 88675123;
function random() { // See http://stackoverflow.com/a/6275875
    var t;
    t = x ^ (x << 11);
    x = y; y = z; z = w;
    return (w = w ^ (w >> 19) ^ (t ^ (t >> 8)))/(4294967296/2);
}

In what way will the output be different to Math.random()? Also, why does 2^32, 4294967296, need to be divided by 2?

Answer 1

In what way will the output be different to Math.random()?

Math.random uses an implicit engine specific algorithm which has no seeding capability, whereas your implementation is platform independent and explicitly seeded.

Also, why does 2^32, 4294967296, need to be divided by 2?

It is used to ensure a multiple of two, so the random number generator does not fail on an odd number.

References

• JavaScript Random Numbers
• JavaScript Crypto
• Better Random Numbers for JavaScript
• Some Bugs in JavaScript Implementations
• Node Alea PRNG Module
• curand generator fails on odd number of elements