Slashdot Mirror
Office Surveillance: Locating And Tracking 802.11b
securitas writes "The NY Times recently ran an article about locating and tracking users of 802.11b WiFi networks in three dimensions using triangulation (Google) with multiple base stations. The goal is to create context-aware networks that can allocate bandwidth and provide location-based services such as uploading relevant information to a PDA. The article can be seen in a new light when coupled with the growth in workplace surveillance of employees by corporate executives (Google / short version at IHT) and the associated practical, ethical and legal problems. Interlink Networks 802.11 wireless detection and tracking white paper (PDF)." (This seems as good a place as any to mention Kensington's handheld 802.11 detector; they claim it to be the only such device on the market today. This is the cheapest detector I've seen; have the others all disappeared?)
The signal gets weaker as it passes through walls. Therefore, the signal strength can not be easilly be correlated to a distance from the base station for purposes of triangulation.
Triangulation traditionally relies on measuring distance through signal strengths and so is limited to an outdoor environment, where the signal loss per kilometer can be predicted with much greater accuracy than in an indoor environment.
The article is short on technical details -- did they somehow also enter a 3D-model of how the building weakens radio signals, and use that in order to create three 3D-shapes at the point of intersection the transmitter can be located? Just like traditional triangulation, but with weirder shapes than simple spheres...
Perhaps a better way would be to use "ping" to check the travel times, rather than the signal strength, compensating for any delays imposed by TCP/IP-stacks and hardware etc. Is this even possible, or is the Signal/Noise ratio just too low?
And an LCD display showing ESSID's...
This is a pretty stupid approach from the communications theory point of view. 802.11b frames contain a pretty long preamble in front of the packet header and data payload. This preamble (basically 11-bit barker sequences convolved with a prn-sequence) have excellent autocorrelation characteristics since they must be used for time and frequency syncronization at the RX station.
By cross correlating the received signal with the (known) barker sequence at all three base stations precision would be increased drastically as it would be possible to measure the actual time lag (->way) the signal took to the receiver.
Positioning in office environments using WLAN really isn't that new. Microsoft did it in 2000 with the RADAR system (http://citeseer.nj.nec.com/bahl00radar.html), and loads of people have tried since.
:-(
There are two approaches to it:
1. Use signal strength to estimate range and then multilaterate. This usually does a poor job because you can't match distance reliably to signal strength because of wall attenuation etc. Also, most WLAN systems quantize the signal strength into a few bins.
2. Pattern recognition. Have a calibration phase where you put the device in lots of positions around the office, measuring the signal strengths to various stations. Record all this. Then try to match what you're seeing to this database of strengths to localise yourself. Problem is, the radio environment changes VERY easily, so you need lots of points in calibration. Plus, if the environment changes, so do the signal strengths!
The best I've seen for a WLAN system achieved accuracy to about 2 metres. That used quite a few WLAN dase stations, too. And they had a fair error on that too - enough that you wouldn't be able to guarantee which office you're in...
Location indoors is a tricky business. It's an active research area. The best so far is based on ultrasonics (the Bat system at (www.uk.research.att.com/bat). UWB looks good too (www.ubisense.net).
It occurs to me that this system could seal a major hole in the concept of wireless security. As we all know, the biggest problem with trying to lock down a wireless network is that it's basically just a radio broadcast and anyone within range can easily tap into the signal (whether they can get anywhere from there is another matter, but theoretically it's always possible to crack through software guards). But if the triangulation worked well enough, then a system could be set up to, say, detect if a client is sitting on the ground in the alley next to the building, and if so shut off the connection to that client. Or it could be used to limit wireless access to only clients in certain offices or floors - no access for random people in the lobby, for instance.
Karma: Chameleon (mostly affected when you come and go, you come and go)
Following the links, Kensington doesn't list an MSRP or sell it directly, but the other links indicated the "going rate" for the toy is $22.00, and I think that's well within budget for a computer toy.
It could really use an external antenna though. If it had this, (or if the unit itself exhibited some amount of directional reception?) then it would be much more useful to find the actual location (down to say, which building on the storefront) the hotspot was at. The closer bench gets the better connectivity!
Maybe someone will post a hack shortly that shows how to jurry-rig an antenna port on the little bugger. I'd also like to "me too" a previous post that suggested an external power connection. Just keep the puppy sitting on your dash whilst driving around town until the green lights start climbing up.
Was anyone able to spot where these could be bought at? (this really looks like something ThinkGeek would carry)
I work for the Department of Redundancy Department.
One pertinant thing I noted in the article was the following:
As part of that work, Dr. Junglas modified a Wi-Fi network that operated in the business school's two buildings so that each of its many base stations had a radius of about 15 feet.
Emphasis mine. This is an insanely dense network of AP's! At over $100 a pop for a cheap one, it seems wildly impractical to simply use stock access points with software corelation to figure out where people are - assuming such density is required.
In a commercial deployment, AP's are going to be deployed in such a way as to give good coverage without costing too damn much. ie: as few AP's as will give adequate coverage for the site.
There are other solutions, of course. Using a phased array antenna (sorry, no cool rotating dish) to get a direction and using signal strength to approximate range (random attenuation in the site will have a large affect on accuracy) or using multiple antennas in fixed locations to triangulate a source location (the more vectors you can get, the more accurate your fix will be) Using signal timing between different AP's (time difference between arriving signals) is plausible, but would add considerably to the cost (current AP's aren't equipped with ultra accurate clocks and transmission times over the network aren't accurate enough for the purpose.)
Phased arrays for direction finding use precise measurements between antenna elements to get their accuracy. They effectively use a harmonic tone to determing the shift angle between antennas, and thus the relative direction to the source. Accurately placing and orienting the AP's would be vital.
Locating wireless source points isn't exceptionally hard, and could be rather useful. But accuracy costs. Existing AP's would give limited accuracy, so this study used lots of them. More acurate location capability on an AP would cost more.
Take your pick.
Never attribute to malice what can as easily be the result of incompetence...