Today I like to introduce a new Pseudo Random Number Generator which I have invented over the last Weeks. Any constructive Critic or Comment is welcome.

With a 32-bit Seed passed through it generates 32-bit Random Values, with minor Tweaks even 64-bit or perhaps 128-bit.

The Results of xqRNG32 as 32-bit binary Output passes all empirical and statistical Tests for Randomness as there are FIPS-140-1, Diehard Test Battery, Frequency-, Poker-, Runs-, Long-Runs and Serial-Test, also Monte Carlo Value of Pi, Arithmetic Mean, Serial Correlation Coefficient and it generate a good Uniform Distribution of Zeros (0) and Ones(1).

Also it passes the complete "TestU01 test battery", namely SmallCrush, Crush, BigCrush, Rabbit and Alphabit without showing any kind of recognisable weakness.

The Idea behind is "porting" a mechanical Lottery Machine into a simple and fast Computer based Algorithm. So imagine there are 256 Tubes made of Metal, all of them lined up in a vertical Way one close to the other in a long Row, filled with numbered Table Tennis Balls.

The functionality of the imaginary mechanical Lottery Machine would be as follows:

#) Each Metal Tube will be filled with 2^32 Table Tennis Balls in a randomly fashion, while each Ball carries a printed HEX-Number from 0x00 up to 0xFF.

#) Horizontal beneath the Row of Tubes is a "Selector", carrying one Glass which has the exact same Diameter as one Tube.

#) This Glass will now be constantly moved from left to right on the "Selector" under those Tubes until it comes to hold under one randomly selected Tube.

#) From this Tube one Ball will drop into the Glass and thereafter will be placed on a Tray in Front of the Lottery Machine.

#) After another 3 Times repeating this Procedure there are now 4 Balls with HEX-Numbers on the Tray.

#) These concatenated 4 HEX-Numbers build the 32-bit Random Value as the Result.



The C++ Listing of the imaginary "Lottery Machine" will read as follows:

Code:
#---------------------------------------------------------------------
/* xqRNG32, Pseudo Random Number Generator based on pqRNG */
/* written by Karl-Uwe Frank, Copyright (c) 2011 Adverteam Limited (UK) */

/* 
 Free to use with or without modification and without a fee is granted 
 only for private, research, academic or other non-commercial purposes. 
 http://www.adverteam.co.uk/xqRNG32/LICENCE
*/
 
static uint32_t seed, Q, P, R, X[256], aP[256], aR[256];

int xqRNG32_Init() {
	int i;

	Q = seed ^ 0x11B923B;
	P = seed ^ 0x6FC55 + 0x1754;
	while ((P & 0x7) != 3) P++;
	R = (P >> 3);
	while ((R & 0x7) != 5) R++;

	for (i=0; i<256; i++) {
		Q = ((Q ^ R) * P);
		X[i] = seed ^ ((Q ^ R) * P);
		aP[i] = Q;
		while ((aP[i] & 0x7) != 3) aP[i]++;	
		aR[i] = (aP[i] >> 3);
		while ((aR[i] & 0x7) != 5) aR[i]++;
	}

	Q = seed;
	Q = ((Q ^ R) * P);
}


unsigned int xqRNG32() {
	unsigned char Tube, ByteX[4];
	int i;
	
	for (i=0; i<4; i++) {
		Q = ((Q ^ R) * P);
		Tube = Q >> 24;
		X[Tube] = (X[Tube] ^ aR[Tube]) * aP[Tube];
		ByteX[i] = X[Tube] >> 24;
	}
	
	return (ByteX[0] << 24 | ByteX[1] << 16 | ByteX[2] << 8 | ByteX[3]);
}
#---------------------------------------------------------------------

Annotation: This PRNG should not be considered cryptographically secure.

Further Information, Test Results and a documented C++ Source Code of xqRNG32 can be found at http://www.adverteam.co.uk/xqRNG32/

Cheers,
Karl-Uwe

---
Copyright (c) 2011 Adverteam Limited (UK), Karl-Uwe Frank
All rights reserved.