3 ms·
I don't understand why nobody ever mentions CCD noise as a source of quality randomness. Billions of devices, now, including all phones and most laptops, have a
by ncmncm 7y ago
I don't understand why nobody ever mentions CCD noise as a source of quality randomness. Billions of devices, now, including all phones and most laptops, have an extremely productive seed source with many millions of pixels, each pixel supplying at least a quarter bit of entropy. It works equally well with a sticker over the lens.
Microphones are also good sources, down in the low bits, because true silence does not exist. Like CCDs, even if you don't think much of the entropy in any one sample, there are one hell of a lot of samples to draw on.
If you have an accelerometer, that's another six sources. If the device is on the table, jiggling the vibration motor provides enough activity.
The RDRAND behavior in that shipment of Zen2 chips suggests that the hardware RNG can be turned off by the System Management backdoor device, and that a code path to do it is active and (a little bit too) available. People messing with CCD pixel readings seems less likely.
- usr1106 7y agoThe problem that lead to the Linux 5.3 changes was early boot. Even if your system has a camera early in boot it is not likely to be initialized. I fear it might be difficult if not impossible to have a generic kernel that gets entropy from an unknown model of camera early in boot.
- ncmncm 7y agoGood point. Where to get randomness when nothing is initialized yet is a tricky matter. I wonder if RAM could be mistuned to generate enough randomness. Certainly not always and everywhere, but probably usually, if you are running on bare metal. If not, whoever is has plenty of randomness to hand over.
- basilgohar 7y agoI'm not sure if this is just a typo/mistake, and perhaps what was originally intended, but CCDs are almost no longer present on most consumer devices. The image sensors used now are most commonly CMOS-based, not CCD. What, if any, change to your original assertion that makes, I don't know. That depends on if the quality of the randomness you cited is significantly different between CCD-based sensors and CMOS ones.
- ncmncm 7y agoIt is true that I missed the shift from CCD to CMOS cell tech, but it doesn't change anything. It might be that the camera chip filters out much of the noise, which would be too bad, but there is plenty left.
- chrisseaton 7y ago> I don't understand why nobody ever mentions... Most of these things aren't present in the context of a server, let alone a virtualised one. Nobody's there to literally physically jiggle your virtualised server while it boots.
- Tomte 7y ago> Most of these things aren't present in the context of a server That's a good reason why it cannot be the main source of randomness, but I find parent's point extremely good. Of course, there might be hesitation about activating a camera for "extraneous" purposes.
- ncmncm 7y agoA virtualized server has no reliable source of randomness at all, besides whatever the host provides. They are out of the picture entirely.
- chrisseaton 7y ago> They are out of the picture entirely. They’re a central topic of this article and where the problem we're discussing is usually felt.