New technique produces real randomness | Hacker News Reader

New technique produces real randomness

sciencenews.org

27 pointsfraqed15 comments

There was a previous discussion here not long ago (11 days ago, 159 comments, 385 points): https://news.ycombinator.com/item?id=11719543

Interesting idea but I wonder why this technique is stated new since it looks very easy and comprehensive to me just by reading some sentences. Is this really new or is this just a bloated message containing something that isn't that hot?

Another side note: the article contains very much sentences (more than 2/3) not directly explaining the new technique. The read was not that great for me, and I guess this counts for most of the readers since it should be well known that random numbers are a hard thing. Don't get me wrong. Explaining things is great but the title doesn't match the most of the content.

> But this still falls short of being truly random. If the mouse is on the left of the screen one moment, it’s less likely to be all the way on the right in the following instant.

Except... that's not how it works... Entropy from input devices is gathered from the timing of interrupts. It's not just looking at the actual values.

Clickbait. The algorithm is a new technique for combining entropy sources.

I really don't understand how this is useful. If you have several sources of randomness, even if they are weak, can you not run them through a cryptographically strong hash function and get a more random output?

The voting system sounds even weirder. How does taking groups of threes solve malicious voters? What even are malicious voters?

> The voting system sounds even weirder. How does taking groups of threes solve malicious voters? What even are malicious voters?

This is normally talked about in terms of biased coin tosses. You know Von Neumann's de-skew technique?

http://pit-claudel.fr/clement/blog/generating-uniformly-rand...

> If independence of successive tosses is assumed, we can reconstruct a 50-50 chance out of even a badly biased coin by tossing twice. If we get heads-heads or tails-tails, we reject the tosses and try again. If we get heads-tails (or tails-heads), we accept the result as heads (or tails).

I'm not sure why the submitted article is any different from this technique.

Ok that makes sense for random numbers, but I'm not sure how it would apply to a real world election.

BTW, there is no such thing as a biased coin, only a biased flipping mechanism. You can't inherently bias a flipped coin like you can bias a die.

You can create a biased 2-state generator , but not with a flipped coin.

Yes, but these are mathematical coins not real world objects.

(I upvoted your post. I don't think it deserved the downvote(s)).

I thought hash functions only removed entropy?

I guess it depends on if the hash method has less output than input?

The only way that would make sense is if your hash function is random. That's typically not how hash functions are built, since it would largely render them useless. A cryptographically strong hash function only tries to guarantee that there is no predictable order to the hashes, not that the hashes are random.

Well of course they aren't random, it's a deterministic function. But the algorithm proposed is also a deterministic function. Both claim to produce scrambled outputs from semi random inputs. But it's not clear why this method is superior.

I think the idea here is that you can take two weak sources of randomness and combine them into a stronger, single source of randomness in a homogenous way that doesn't reveal the original weaknesses. Thus it's not adding randomness from nowhere, but composing it in a useful way.

I don't actually know, but that's the only interpretation I could come up with that makes sense to me.

Detecting duplicates is not a solved problem; it's hard. You have to compare strings and everything.

We detached this comment from https://news.ycombinator.com/item?id=11797275 and marked it off-topic.

Reply on news.ycombinator.com