Testimonials

I study the life-cycle of viruses, and I perform lots of tissue culture experiments. In order to try to develop theories to explain some results I was getting, I wrote a computer program that uses a Monte Carlo scheme to simulate infection of cells by viruses. I need a different random number for each simulated virus, in order to randomly assign it to a cell that it ‘infects.’ In order for the results to be meaningful, I need to simulate tens of thousands of ‘cells’ and hundreds of thousands of ‘viruses,’ so I need hundreds of thousands of random numbers. The pseudo-random numbers produced by the Apple Macintosh built-in linear congruental generator proved themselves to be not good enough for the job, as I found that some numbers were chosen too often, a definite no-no for my purposes. Then I saw the NY Times article about this site and gave it a try. First I tried using Random.org numbers to seed the Macintosh generator at frequent intervals during the execution of the simulation, but it did not solve the problem. So I tested using all numbers from this site and they passed my quality test. So now I download several batches at a time of 10,000 numbers between 1 and 40,000 and string them into big files as the sources of my numbers. I'd like to be able to download them in even bigger batches, though. Thanks for a truly useful service!

—David N. Levy, University of Alabama at Birmingham

I discovered Random.org due to the New York Times article on random numbers today. I've already downloaded the three pre-packaged 10 MB files and wish there were more of them (at least three more 10 MB files). I'm using them as audio—interpreted as 16-bit WAV files, they form perfect white noise, which has many uses in acoustics and audio-equipment testing, which is my field. Used in pairs, they form perfect, uncorrelated stereo white noise.

I've been able to get more use out of the first three 10 MB files by reversing their byte order (the resulting white noise sounds the same) and by using various other audio-editing tricks like concatenating the files to produce long streams). I've also used 1, 2, 3 or 4 bytes at a time to produce different audio wordlengths. Thanks to the 2's complement number system, this latter scheme is particularly effective for audio since you always get equal distributions of data points above and below zero.

Your files produce better noise than some pseudo-random schemes I've tried, since the latter can produce an audibly detectable cyclic effects in the sound quality if the sequence length is too short. The ear is an extremely good detector of such patterns. A quick-and-dirty one-time-pad scheme would involve Xor-ing your random bytes with the lower bytes of each 16-bit word on a commercial audio CD to produce the random number table. The recipient would only need your file and another copy of the audio CD. To crack it you'd have to search through every data sample on every CD ever released!

—David Ranada, Technical Editor, Sound & Vision Magazine