Airshark is an experimental application bringing detect-and-avoid frequency-hopping to previously dumb Wi-Fi cards, without the addition of any new hardware.
Created by three boffins at the University of Wisconsin, Airshark uses the limited intelligence of existing Wi-Fi cards to spot other protocols hanging out in the 2.4GHz band. That includes Zigbee, Bluetooth and microwave ovens as well as everyone else, potentially letting a Wi-Fi router switch channels to somewhere a little less crowded.
The software presented (PDF, starts out easy, then becomes highly technical) takes advantage of the increasing intelligence of Wi-Fi cards to sample the signal strength within a channel and search for interference corresponding to known users of the band. When it spots something, such as a cordless telephone or XBox controller, it takes action to mitigate the problem caused by that kind of device, resulting in a faster and more stable Wi-Fi connection on existing hardware.
The latest generation of Wi-Fi routers can already detect other Wi-Fi users, scanning the Wi-Fi channels to find an empty slot, but such techniques are useless when the band is filled with other protocols which are a lot harder to spot. Bluetooth, for example, leaps around the place occupying one-twentieth of a Wi-Fi channel for a millisecond or two, while a microwave blunders across multiple channels but is equally undetectable to a device that is looking for other Wi-Fi users to avoid.
Airshark takes advantage of features exposed by the latest generation of Wi-Fi chips, exemplified by the Atheros AR9280 AGN wireless card, to sample signals within the band used by a Wi-Fi channel. Those samples are checked for recognised patterns to see if known protocols can be spotted and mitigated against.
So, for example, if an analogue cordless phone is spotted taking up a 1MHz-wide channel then the Wi-Fi protocol can use channel-width adaption to squeeze itself into half its channel and avoid the interference, periodically checking back to see with the phone call is finished.
This kind of thing is already possible, but only using spectrum analysis hardware that would be prohibitively expensive to put into a Wi-Fi router. What's interesting here is the ability to do the same thing with standard hardware, and with implications well beyond Wi-Fi.
One of the problems that white space users will face is how to detect other protocols operating in the same frequency bands – to avoid mutual interference. Detecting devices running the same protocol is generally quite easy, but what has been demonstrated here is commodity kit being used to detect (and identify) different spectrum users with no more than a software upgrade. ®