## Webcam Random Number Generation

A couple weeks ago, I purchased a lava lamp for \$5 at a thrift store. It was in brand spanking new condition, and worked like a charm. The only thing going through my head at the time? I can't wait to point my webcam at it, and start generating some random numbers! Okay, well that, and mood lighting for the wife.

Anyway, I wrote a quickie Python script which will capture a frame from the webcam, hash it with a keyed BLAKE2, and output the result to a FIFO file to be processed. The BLAKE2 digest of the frame also becomes the key for the next BLAKE2 instance, making this script very CBC-like in execution (the first function is keyed from /dev/urandom, and each digest keys the next iteration).

 12345678910111213141516171819202122232425262728293031323334353637383940414243444546474849505152535455 #!/usr/bin/python # Create true random seeds (near as we can tell) with your webcam. # # This script will use your webcam pointed at a source of entropy, keyed with # random data from the OS CSPRNG. You could point the camera at: # #   * Lava lamps #   * Plasma globes #   * Double pendulums #   * Rayleigh-Benard convection #   * Brownian motion # # Performance is ~ 2 KiB/s. # Requires pyblake2: https://pypi.python.org/pypi/pyblake2 # # Released to the public domain. import os import cv2 import pyblake2 cap = cv2.VideoCapture(0) webcamfile = '/tmp/webcamfile.fifo' key = os.urandom(64) try:     os.mkfifo(webcamfile) except OSError, e:     print "Cannot create FIFO: {0}".format(e) else:     fifo = open(webcamfile, 'w+') while True:     ret, frame = cap.read()     if not ret:         break     b2sum = pyblake2.blake2b(key)     b2sum.update(frame)     digest = b2sum.digest()     key = digest     fifo.write(digest)     fifo.flush()     cv2.imshow('webcamlamp', frame)     k = cv2.waitKey(1) & 0xFF     if k == 27:         break fifo.close() os.remove(webcamfile) cap.release() cv2.destroyAllWindows()

As you'll notice in the code, you should point your webcam at a source of either chaotic randomness, like a lava lamp, or quantum randomness, like a plasma globe. Because the frame is whitened with a keyed BLAKE2, it could be considered as a true random number generator, or you could use it as a seed for a cryptographically secure pseudorandom number generator, such as those shipped with modern operating systems. If you do use this as a TRNG, realize that it's slow- it only operates at about 2 KiBps.

Here is a screenshot of the webcam itself looking at a USB desk plasma globe, that you can purchase of ThinkGeek for \$10.

The data is sent to a FIFO in /tmp/. If you don't do anything with the data, and let the buffer fill, the script will hang, until you read data out of the FIFO. As such, you could do something like this to reseed your CSPRNG (of course, it's not increasing the entropy estimate, just reseeding the generator):

`\$ < /tmp/webcamrng.fifo > /dev/random`

Lava lamps and plasma globes are only the beginning. Anything quantum or chaotic that can be visually observed also works. Things like:

• Double pendulums
• Brownian motion
• Rayleigh-Benard convection
• CCD noise from the webcam itself
• A bouncing ball on a sinusoidal vibrating table

So, there you have it. Plasma globes and lava lamps providing sufficiently random data via a webcam, either to be used as a secret seed, or as a TRNG itself. Any other systems that could be used to point a webcam at, or suggestions for improvement in the Python code, let me know in the comments.