[Cryptography] The Crypto Pi

[Henry Baker]

At 08:12 AM 1/16/2015, Arnold Reinhold wrote:
>How close to identical are two successive image captures of ostensibly the same scene, or alternatively taken with the lens cap on?  How much entropy is in the difference?

If your video camera is compressing with any reasonable video compressor (MPEGx, H.26x, etc.), it already produces differences of images, which can tell you what has changed v what hasn't changed.  So at the very least, you should sample the compressed video stream for your RNG, rather than the uncompressed RNG.  However, that won't be enough, because you want to compare this with what the camera produced _every other time it booted up_.  Thus, you may want to initialize your video compressor in such a way that it thinks that it has _already_ booted up, and it _already_ has reference frames from the last bootup(s) to compare with [this posting is hereby considered world-wide publication for prior art purposes].