EE Times describes some research out of the University of Illinois by professor Paul Kwiat and doctoral candidate Michael Wayne. I wrote about the value of quantum random number generators and the problems with their classical counterparts earlier.
Kwiat is claiming that this technique is faster than anything else out there. It is based on an optical system, specifically the measurement thereof. As the article explains, the delay between measured photons is guaranteed to be random though it fails to further explain why. I imagine the emitter is governed by some quantum property or they are taking advantage of such a property within the detector. They claim 100 Mbits per second with their current setup and expect to achieve 10 Gbits with further refinement.