I didn't understand "looking for wifi" as a functional criterion of exclusion 
or inclusion.

Do you mean that your friend does not wish to count folks who are jogging over 
to the happy wifi park to check an email and then jogging back to sad non-wifi 
non-park land?

Or, are you implying that seeking SSIDs causes your friend to receive a hit? If 
the latter, that's just bad data collection in that case. Look at tcp flows 
over a minimum of 1 minute, 5 minutes, whatever your threshold is. That should 
weed out folks looking for wifi.

Now, the far more intriguing question here is the percentage of visitors that 
wifi presence can tell you.

There are some hard-core ways of figuring this out, including flying cheap 
drones over the park, counting heads, correlating against wifi data, doing this 
for a week, and then extrapolating a model based on time of day, weather 
conditions (e.g. rain may affect wifi usage because Uber usage goes up), 
testing that model against 50% of your dataset that you've held in reserve, 
refining, etc.

Other solutions could include some volunteers or non-witting volunteers. So, 
gamify this problem, and generate a cheap Pokémon Go knock-off that allows 
people to bag aliens from Wifitar VS Nophoneland. Wifitar aliens are simply 
anyone using a phone with some goofy AR overlays on top to make it fun for the 
whole family, and similar for aliens from Nophoneland, except obviously no 
phone is used. Use the game as a way of increasing publicity for the park, 
discount any game players from your model since they will artificially increase 
your wifi numbers, and then correlate against again half of your dataset you've 
kept in reserve.

Other things that come to mind is surveying folks about their mobile phone 
usage. From such a survey, you may be able to do basic head counting, as well 
as correlation against known numbers of phone usage. Once you know the delta 
between those two numbers, you could use it to inform some coefficients for 
estimating head count.

Now, if you wish to get a bit invasive, I would say to go grab a few $20 
software defined radios, or SDRs, and then count all mobile devices in the 
area, wifi or not. You can use RSSI values, especially if you have multiple 
radios, to even triangulate position, and then you can count all cellular 
traffic signals in the area. You then can correlate against wifi hits from 
above, and now you know the percentage of folks that have wifi turned on VS all 
folks with cellular devices. You can then use other datasets to learn what 
percentage of the public leaves wifi on or even has a cell phone, and then use 
that data to appropriately scale your own numbers up/down.

Interesting aside, this is how the US government listens in on mobile phone 
calls all the time. the device that law enforcement uses is called a Stingray, 
and it pretends to be a cell phone tower, so all the phones hook on to it, and 
then they can man-in-the-middle attack the voice and data traffic going through.

This is now trivial to do yourself since gsm encryption has been cracked and 
openly known for a while (well, trivial other than the whole prison thing, 
violating people's privacy thing, and basically being a super evil person).

Other options that come to mind are making sure to have your routers possibly 
log collision avoidance on the 2.4GHZ spectrum. People may leave Bluetooth on 
more than they leave wifi on, especially now that Apple has no headphone jack, 
so you could use this as a possible additional way of counting mobile devices, 
then counting wifi and doing that delta from above.

Ok, I capped myself to 3 minutes to respond to this, because I'm totally 
procrastinating on paperwork, but I'm happy to discuss. the above ideas are far 
less implausible than they may seem. I used some of the approaches above to 
write in-door location algorithms back before it was ever cool. it was a 
startup in the 2007 era for RTLS in hospitals. Cisco threw millions of dollars 
at the problem and only had 60-foot spherical accuracy (which was cute). Our 
approach achieved room-level. So, this stuff is doable.

It really comes down to resources and comfort level. This is a tractable 
problem, it seems, but some solutions are rather frowned upon.

Back to paperwork. This was way more fun!

Take care,

President, Prime Access Consulting, Inc.
Twitter: @SinaBahram
Company Website: http://www.pac.bz
Personal Website: http://www.sinabahram.com
Blog: http://blog.sinabahram.com

-----Original Message-----
From: mcn-l-boun...@mcn.edu [mailto:mcn-l-boun...@mcn.edu] On Behalf Of Nik 
Sent: Thursday, December 01, 2016 6:22 PM
To: mcn-l@mcn.edu
Subject: [MCN-L] Tapping the MCN Brain Trust

I have a friend who runs a large, free public-access wifi network in a park. 
The network requires no authentication. There is modest promotion of the 
availability of free-wifi. He’s looking to estimate the total number of 
visitors to the park from the number of unique clients he sees on his wifi 
network. Despite the fact that a significant proportion of visitors have their 
smartphone with them, only a certain percentage will appear on the network due 
to a variety of factors including phone settings and a user checking to see 
whether there’s wifi available.

What percentage of the total visitor number does the MCN brain trust think he 
will see on his network? Or maybe put another way, what percentage of the 
population looks for free wifi?


Nik Honeysett | Chief Executive Officer

M (805) 402-3326  P (619) 331-1974  E nhoneys...@bpoc.org 
2131 Pan American Plaza, San Diego, CA 92101

A technology collaboration that connects audiences to art, culture and science.

You are currently subscribed to mcn-l, the listserv of the Museum Computer 
Network (http://www.mcn.edu)

To post to this list, send messages to: mcn-l@mcn.edu

To unsubscribe or change mcn-l delivery options visit:

The MCN-L archives can be found at:

Reply via email to