[pgn674]

I thought that pseudo random referred to the way that the seed number was generated. Like, by sampling disk I/O metrics, thermometer readings, and mouse movements and combining them together. Pretty random, but still predictable in certain ways. If you wanted a truly random seed, you'd need to sample something the physicists say is actually random, such as cosmic microwave background radiation or certain subatomic processes following the uncertainty principle.

Did I learn wrong? Are all random numbers derived from an appropriate recursive function considered pseudo random, no matter how the seed was derived?

[quercusa]

There are true RNGs that use a physical source of randomness such as resistor noise [0] feeding into a comparator. Most of the things like mouse movements can be random(ized) but tend to be really slow (few random bits per second).

Psuedo-RNGs typically use an algorithm (such as a hash) which is very hard to reverse. The next value is deterministic and easy to predict if you know the algorithm and input values, but effectively impossible (i.e., 'random') if not.

[0] https://news.ycombinator.com/item?id=27447766

[lann]

> Psuedo-RNGs typically use an algorithm (such as a hash) which is very hard to reverse.

Note that this is only necessarily a property of Cryptographically Secure PRNGs. Some PRNGs are not designed to be difficult to predict at all.