▲ 48 points
back
17 comments
But are they more random than a wall of lava lamps?
https://en.wikipedia.org/wiki/Lavarand
In all seriousness, the movement of large objects would seem inherently more random than the movements of individual particles. The actual number and real points of failure come not from the objects but from the devices that measure their position.
> the movement of large objects would seem inherently more random than the movements of individual particles
sounds contrary to law of large numbers
Being small means you can integrate it in the CPU die. This is very important for SecureBoot/DRM kind of things, because it means you can't easily manipulate the secure processor entropy.
I know HN doesn't like SecureBoot or DRM, but this kind of stuff guards against maid attacks.
For example with a lava-lamp someone could replace your webcam with a hacked one that manipulates the picture in a subtle way that removes entropy. But you can't replace a SecureBoot CPU, it will be noticed if the TPM lives inside the CPU.
A ring oscillator circuit is small and fast too. I'd be curious to hear how this new idea compares to the contemporary approach.
It was my understating that the randomness primarily comes from sensor noise and not from the lava lamps themselves.
https://blog.cloudflare.com/lavarand-in-production-the-nitty...
Sensor noise absolutely part of it but if that were the only source then pointing the camera at a fixed image would be fine. In this case, the combination of the randomness of the lamps plus that of the sensors noise makes it more difficult for an attacker (just taking a picture of the same lamps at the same time won't be enough).
Pointing the camera at a fixed image would, in fact, be fine.
The lava lamp is just for publicity. Any old camera attached to any old gadget provides more truly random bits than you will ever need, no matter where it is pointed, or isn't. A microphone, likewise. A radio receiver, likewise.
People rolling dice to get random numbers is the height of silliness. Using a specially built gadget is about equally silly: you never know if you built it right. AMD shipped microprocessors with a "true random number generator" device, and instruction to access it, that gave wholly non-random numbers. They had to distribute a microcode patch for that, demonstrating that microcode is able to deny you those supposedly truly-random bits any time the spooks like.
With PCI cards like from Id Quantique [1] which generates true random numbers using quantum optics, isn't true random number generation a solved problem?
Guessing the quantum mechanics predicts true random numbers, but the messiness of real-world implementation makes it likely to be not so "true" after all.
[1]: https://www.idquantique.com/random-number-generation/product...
I won't comment on whether it's a solved problem in general, as I don't actively follow the area very closely, but that particular QRNG has been shown to be more biased than /dev/urandom [0], and other QRNGs have exhibited similar issues.
There’s room for another solution, especially if it could someday lead to a product sold for under $1500.
A cellphone camera is, what, $5? Maybe you already have one. Or an audio input jack you can crank up the gain on. Maybe a radio transceiver, like for wifi or bluetooth?
All our gadgets already have sources for way more truly random bits than we could ever find a use for.
If you feel you really need to build something from scratch, resistor thermal noise is random. Likewise, diode shot noise. Total parts cost for either is well under $1. Combine them if you like; you are still well under $1.
Why does this have to be so complicated. Serious question, could you point a CCD with dirty lens in front of it and point it towards the trees moving randomly all day from wind ?
I'm no expert in cryptography, but my thought is that it isn't enough just to be random. For cryptography, you need a source of randomness (unpredictable) that cannot be easily manipulated (which could make it predictable to the manipulator).
In your example, if I could locate your camera I could put a video I control in front. Now I can direct the outputs to be more predictable to me.
First I’m hearing about skyrmions. Fascinating stuff. Could this be harnessed in a small form factor?
Skyrmions were known in theory for a long time, but only experimentally realized less than ten years ago. I guess you can say they fall within the larger field of spintronics, where people seek to exploit spin-charge coupling.
The most well-known example is of course giant magnetoresistance (GMR), and later tunnel magnetoresistance, which was a key piece of harddrive read head technology development in the late 90s and 2000s.
GMR was discovered in 1988, and received the 2007 Nobel prize in physics, and it was actually miniaturized sufficiently to be used commercially in just 10 years.
> When the skyrmion size changes, the voltage changes to an extent that is easily measured
I'm a bit excited! I wonder what easy means here; measuring the voltage, generating the 2D material film, keeping it at the right temprature. 10 megabits sounds sweet, but I believe you could get that from CPUs at least 10 years back. Hope they can make this small and energy efficient, time will tell!
Random number generators are so cool!
They're also mostly a solved problem though. Entropy is everywhere and modern distillers seem to be pretty low risk, now that we're in an era where. industry standard crypto rarely gets broken at the primitive level, as far as we know.