# How to produce random numbers from seismic data?

I have read that the University of Chile uses seismic data to produce random numbers. (https://random.uchile.cl/randomness-beacon). Does seismic data real have randomness and if it so, which information is used in seismic data: magnitude, time, vb.?

This quote is from the article:

Use of Local Entropy: after collecting data from the external sources, it is necessary to generate entropy locally. For that, we use our own TRNG (True Random Number Generator), a piece of hardware that produces random bits by relying on a quantum process on an electrical circuit. At every minute, the TRNG generates a 512 bit value.

I found the article confusing. They allude to seismic data being used but the quote from the article suggests that the random number is generated using hardware that generates the random number based on quantum mechanical processes.

The article also mentions the use of cryptography. Perhaps the daily seismic data (which is essentially unpredictable), is used as a seed for the cryptography.

Any unpredictable source that can contribute to numerical entropy is a good source for trying to generate truly unpredictable random numbers. People even use the somewhat unpredictable timing of how long it takes to access data on a disk drive as a source of randomness. Generating truly random numbers is an impossible task. That does not mean that /dev/random and all of the sources that feed it are not a good start. They are. Similar concepts exist on other operating systems. In general, the more entropy the better.

• That said, computers cannot generate an infinite sequence of numbers that is truly random. This falls into the category of things that cannot be computed by a Turing machine. Commented Jan 21, 2023 at 15:12
• It sounds like the output of the TRNG (True Random Number Generator) needs random input. Commented Jan 23, 2023 at 2:35
• @SteveSaban An infinite amount of random input is needed to create a true random number sequence. What is done instead (at least for now) is to gather enough random input and use a complex enough algorithm that a code breaker would take far too long to break the code. Cryptographically secure random number generators aren't true random number generators, and that's a potential problem. How much time qualifies as far too long of a time span? Attacking the code with a supercomputer, a huge network of hacked computers, or a quantum computer could potentially reduce the time drastically. Commented Jan 23, 2023 at 12:06
• @SteveSaban If it takes a couple hundred year for the best computers to track your financial transactions, that's probably more than good enough. But if takes only seconds, that's not close to good enough. Your cryptographically secured password / passphrasre and your two factor authorization require a decent amount of unhackable time for security purposes. A small amount of time would leave hackers more than enough room to slip between the cracks and hijack your money away. Commented Jan 23, 2023 at 12:26