Path: csiph.com!eternal-september.org!feeder.eternal-september.org!mx02.eternal-september.org!.POSTED!not-for-mail From: Paul Rubin Newsgroups: comp.lang.forth Subject: Re: Ruby/Forth challenge for WJ Date: Wed, 13 Jan 2016 11:39:39 -0800 Organization: A noiseless patient Spider Lines: 22 Message-ID: <871t9luv44.fsf@nightsong.com> References: <35f4686c-f2c4-4ed8-9c46-2238645b9b97@googlegroups.com> <569528f5$0$23800$e4fe514c@news.xs4all.nl> <2016Jan13.164504@mips.complang.tuwien.ac.at> Mime-Version: 1.0 Content-Type: text/plain Injection-Info: mx02.eternal-september.org; posting-host="89fc4f87faf042e7e4d428a3639a51e3"; logging-data="17002"; mail-complaints-to="abuse@eternal-september.org"; posting-account="U2FsdGVkX1/d3bHCWofoQbDhx2UIhLAR" User-Agent: Gnus/5.13 (Gnus v5.13) Emacs/24.3 (gnu/linux) Cancel-Lock: sha1:Kp0im+wO1Rf6ruyz907cWKdbveo= sha1:Wl5nGxPMtPbgpS5mtHwE5B4Bk6E= Xref: csiph.com comp.lang.forth:44834 rickman writes: >> With a 20 Megapixel camera, if you only assume 4 bits of randomness >> per pixel, you can get 10MB of randomness per shot... You need to >> post-condition the data so that, e.g., dead pixels or dust on the >> sensor don't compromise your random number generator. > Has anyone tried this? I can see huge issues relating to variations > among the pixels which would need to be averaged out some way. That digicam performance is limited by shot noise is well known, though I'm skeptical of getting 4 bits of physical randomness per pixel, especially with small-sensor consumer digicams where only few thousand photons are enough to saturate a pixel.[1] Yes of course you have to massage the output: there's a bunch of mathematical theory ("randomness extraction") about how to do it with guaranteed minimal loss of information, but in practice most people just run the data through a cryptographic hash function. In practice again, if you can get a few hundred bits of physical entropy, that's enough to seed a CSPRNG and get unlimited output that doesn't seem feasible to distinguish from actual random bits. [1] http://clarkvision.com/articles/does.pixel.size.matter/