5
$\begingroup$

I've been looking into the difference between PRNGs and proper RNG techniques. One that I particularly like is the idea of tuning a radio to a certain frequency and bandwidth and just listening to the noise. Every so often (or continuously) pick a sample out of the noise and build a random number out of the samples.

There are a few obvious frequency ranges to avoid for this purpose:

  • Actual AM/FM radio stations (very predictable, not noisy at all)
  • TV stations (as above)
  • Protected RF bands (government, cellular, etc)

Based on noise-generating natural (and digital) phenomena, my question is: what frequency ranges are likely to be the noisiest and thus carry the most entropy for the purpose of a true random number generator?

$\endgroup$
  • 2
    $\begingroup$ Why use a radio? A reverse biased diode will generate plenty of avalanche noise. It needs post-processing but it's a lot better than radio. $\endgroup$ – Nils Pipenbrinck Feb 8 '15 at 14:45
3
$\begingroup$

This is not likely to be a good method to generate random numbers -- at least not on its own, without significant additional analysis. You need a source of random numbers that is not only noisy, but also unpredictable to an adversary. It's not clear that radio will meet that need. If you listen on a FM band, an adversary who is nearby can also listen on the same FM frequency and will likely see something very similar to what you see, so will have a huge leg up in predicting the numbers you generated.

Bottom line: don't try to roll your own random-number source. It's better to use a time-tested scheme, like a crypto-strength hardware RNG, or /dev/urandom, or a similar crypto-strength scheme.

$\endgroup$
  • 1
    $\begingroup$ +1 Besides, using RF noise for random numbers is wide open to active attacks: an attacker can just transmit on the frequency your RNG is tuned to. $\endgroup$ – Ilmari Karonen Feb 9 '15 at 18:07
  • 1
    $\begingroup$ @IlmariKaronen Analogue reception intricacies and the avalanche effect mean that an attacker can never create a specific binary sequence following entropy extraction. He might be able to jam it though, but that really falls within the realm of EMP attacks so therefore irrelevant. $\endgroup$ – Paul Uszak Jun 4 '18 at 20:26
1
$\begingroup$

There isn't one really for the amateur TRNG builder. Have a look at the frequency allocation for 0 - 300 GHz below:-

spectrum

There are no gaps. A common First World problem is the provision of spectrum for all the intended uses. That's why microwave ovens share WiFi spectrum and Freeview TV competes with mobile phones.

What might seem a quiet (or hissy) gap between radio stations just means that you've not picked up anything at that time. Wait till sunset when the Ionosphere is on line, and you might hear France. Similarly when Stan fires up his arc welder next door you might hear that, or a passing police car.

The other significant issue with unpredictable entropy sources is measuring the entropy rate. That's notoriously difficult for a simple fixed source, but much worse if it can vary in frequency and /or amplitude. And radio varies so much, notwithstanding atmospherics. A mistake in assessing the input rate to entropy extraction risks breaking the golden rule of TRNGs (entropy out < entropy in). So there is the risk your DIY TRNG becoming a common pseudo random number generator.

No one uses radio noise as a TRNG for these reasons. I specifically and with prejudice exclude random.org. That's a private organisation with very little hard detail of how they do what they claim to do. And clearly they only have access to exactly the same spectrum as everyone else.

The hiss is actually thermal noise on the radio's detector and pre amp circuits. You can short circuit this and directly exploit thermal /avalanche noise in a reasonably simple custom circuit pulled off the internet. You can get 1000's bps of entropy this way.

$\endgroup$
  • 2
    $\begingroup$ 1420.4 MHz, which is the hydrogen line frequency. You cannot use that for anything because: "My God, it's full of stars " $\endgroup$ – b degnan Jun 7 '18 at 0:58
-2
$\begingroup$

Use the entire RF spectrum to create a number. As long as you can get a variable (zero to one) formula demodulation scheme. Multiply the percent by the encryption spread number.

$\endgroup$
  • 1
    $\begingroup$ All of it? Is that scientifically even possible with one receiver and less than a billion pounds? Plus some frequencies are illegal to monitor. $\endgroup$ – Paul Uszak Jun 4 '18 at 20:29
  • $\begingroup$ @PaulUszak what frequencies are illegal to monitor? $\endgroup$ – Richie Frame Jun 4 '18 at 23:03
  • $\begingroup$ @RichieFrame I won't even mention Their transmissions such as US SATCOM for remote drone operation @ 20 - 30 GHz. So, I'll take taxis. Illegal under Wireless Telegraphy Act 1949. Actually, it's all private radio transmissions in the UK, that was just an example. Monitoring a mobile phone (encrypted or not) would probably get you done as well for telephone interception. And in the current climate, geeks just WiFi war-driving around the City of London would probably get them arrested if not shot. $\endgroup$ – Paul Uszak Jun 5 '18 at 0:01

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy