So, a couple of local stores teamed up and want me to make a product for them.
Basically, they want me to place small wifi sensors everywhere and using the MAC addresses, they want me to make a mapping of how most customers walk through their stores.
I can do this, it really isn't that hard.. however.. is this morally wrong? Would it make me one of those evil cunts collecting private data?
I'm not sure about it to be honest. They also want me to make a system to figure out what type of people the customers are.. female.. male.. pregnant.. teen.. etc.
What do you think?
>>51899975
I don't think it is morally right, but money is money. Using machine learning for the customer data?
>>51899975
NIGGA MONEY IS MONEY YOU GOTTA HUSSLE FOR THAT GREEN DO IT
>>51900092
>>51900006
I isn't all about the money, I don't want to be the next buying-aids-vaccine guy in the newspaper.
>>51900216
you wont, it's not as bad because you're not a millionaire.
All the more reason to switch off wifi when you don't use it.
An interesting, sort of related talk:
https://www.youtube.com/watch?v=ubjuWqUE9wQ
It's not private data if the customers are inside their store on their private property and using the Wi-Fi provided by the same company. If the customers connect to the wireless they are in effect surrendering the right to any and all privacy at that point in time with respect to the Wi-Fi usage.
Learn how the law works, stupid, and don't come to 4chan asking for advice, you're beneath this place and that's pretty fucking low already to start with.
>>51900318
Well, I will be installing nodes everywhere and whereeveer the MAC is the closest to it will be logged, so that is how I make paths. It has nothing to do with wifi from the stores. It has to do with the wifi on the phones, which people leave on all the time sending out mac adresses.
>>51899975
A MAC address is not private information linked to a persons identity. Be sure to take money in installments with stated milestones for each pay out. A system that learns age and gender is a tall order. You'll probably want to give people a RFID membership card that gives discounts.
>>51900405
I would think that tracking customers via cameras would be more efficient. I assumed that big box stores were already doing this.
>>51900405
Technically you could do it like gps does with just 4 nodes and measuring time between packets received by each device
>>51899975
I think someone forgot to take his medication.
>>51900420
I was thinking of placing a sensor next to the door to check the height of the person and based on that label child, teen or grown up. Its not age, but its something.
>>51900442
It's not, The whole idea of this system it wil show how people walk in stores and how they navigate in the mall. Camera systems often arent connected and stores install them them selfs.
>>51900449
Nah, the stores are too big for that.
>>51900420
This.
You could build a customer profile on them if you install a node at each checkout and connect their purchases with their MAC address. Then the next time you log their info, you would know what their purchase history is. This isn't anything new though, loyalty cards have been doing this for ages.
RFID is a possibility, but you would have to put the sensors in such close proximity to the customers that it would be difficult to hide.
Even with the camera tracking, it would be difficult, if not impossible, for you to gauge pertinent info about the customer, barring some FBI-level AI software.
What kind of store
>>51900488
>It's not, The whole idea of this system it wil show how people walk in stores and ho...
I meant that you could install AI software to assign each customer that walks in a value, and then track them through a network of cameras.
>>51900488
Btw what devices will you use for this?
>>51899975
this is already a thing, why dont they just save some money and buy the beacons?
>two fucking companies at the last incubator i was at was working on this
>some asshole asian guy walks in and places some rubber rock looking thing on the table
>you guys realize your product was made years ago right?
they closed up shop pretty fast
>>51900502
The question is NOT how. I know how, with MAC detection.
>>51900505
Two grovery stores, a store will of cheap crap woman like, a gardening store and a huge clothing store like H&M.
>>51900524
Which isn't nice, because that means installing servers in each store for each camera system. With MAC tracking I can make a single netwerk that covers the whole mall, I can even place them on toilets and parkinglots.
>>51900538
Raspbery Pi 2
>>51900551
I checked the current options, they suck.
>>51900579
>I checked the current options, they suck.
some of them have apis
might be cheaper to buy their systems and wire into it (cheaper as in making all the sensors, remember time is money)
>>51900609
If you make one, you make them all.
RPi + software + wifi antanna
for backend:
utp cables
poe switches
dell r210 server
Fuck buying other people's shit, I can make it better myself.
>>51899975
Unless you're installing quite literally thousands of nodes per venue (essentially one for each aisle) and having the stores tag zones (Zone 1 is women's clothing, tag it as a typically-female zone), you're going to have a hard time determining the demographics of a customer from their MAC.
Plus, didn't Apple and Google switch to random MACs for probes to prevent EXACTLY THIS?
If you actually go through with this, OP, please have a terminal where you can sign in to an opt-out network, which would record their MACs and purge and discard data originating from that MAC.
>>51900641
so you're going to buy like 30 raspis and wireless cards?
if he wants actual mapping, you're going to need a lot more than one
that doesnt sound cheap
>>51899975
How different is this from targeted ads/content?
Do it faggot!
>>51900675
The budget is 45 - 55 k, and closer to 300.
>>51900666
I don't think they did? And you are wrong, the only thing they want is to see how people walk through their stores/mall. From entry to departure.
>>51900727
RIP OP
http://arstechnica.com/apple/2014/06/ios8-to-stymie-trackers-and-marketers-with-mac-address-randomization/
desu OP, I'd be really interested in working on a project like this
To check height place a device at entrance and check the length against the floor. If that length is inferior, someone came in. Also, the difference between open and someone there will give you height. Use height to define age.
Where are you from?
You should use triangulation to do this, not the nearest node.
It will be cheaper, more accurate, and only require 3 nodes minimum.
>>51899975
>>51900808
After looking through some papers I had collected, there is a statistical principle called MCMCDA (Markov Chain Monte Carlo Data Association) which tries to construct progressions (I.e. paths through a space) from unconnected data (random MAC addresses). This would require significantly more processing power than simply recording occurrences of MAC addresses, and would be if dubious accuracy, but would give slightly more valuable data.
How do the stores want to ingest the data? A per-store dashboard?
>>51901038
The problem is that the nodes typically have very little idea where they are in the 2D space of the store rendering triangulation ineffective.
>>51901097
The nodes should never move, therefore they should know exactly where they are.
Place 4 nodes in each corner of the store, and have them compare the signal strength and you are able to determine where exactly the signal originates from.
And as you mentioned that it is 2D, you can ignore the Y coordinate making it even easier.
>>51901134
4 nodes, 1 in each corner*
>>51899975
Amateur radio operator with experience in radio direction finding here. The radio spectrum is considered public property. That's why governments regulate it. The MAC address is just a number that identifies a device and therefore a person in this instance.
So I answer you with another question. If it were common for people to walk around shouting social security numbers constantly would it be wrong for you to note where they were walking in your store and what number they were shouting?
If you don't want people to note your social security number you don't go around shouting it. If you don't want radios to note your wifi MAC address you don't turn your transmitter on. That's effectively what a wifi device does when it transmits. It shouts at the top of its lungs a number.
>>51901134
>>51901097
1134 is right, you literally only need 4 rpis and you get accurate data. Just putting a ass ton of rpis will only accomplish one thing, spending an ass ton of money.
Why haven't stores been using IR to do this tracking? set up IR cameras near security cameras and then splice every IR camera together, and you have a heatmap, shit if you wanted to you ould track individual people, only problem would be them going to the bathroom and leaving IR view but you could just throw anything that doesnt come from the entrance out as junk.
>>51900488
There's tonnes of cctv systems that support this sort of stuff and has become a main selling point of video analytics over conventional PIR detection.
Look up something like avigilon or adpro. They all do exactly what you're after
>>51900318
He's not really asking if it's legal. Many things are legal but seem morally wrong to most people.
There Is a clear distinction between law and morality allot of the time, especially when you get large countries with many inept politicians.
My favourite example from Australia is that live odds betting online on sport is illegal. But some companies apps activate the microphone, it does nothing but it is on, and then they claim it's actually the same as calling up.
Technically legal (for now) but morally wrong as people say this will be the pokies for dem youngin's.
>>51901134
If OP's definition of a mall is the same as mine, your proposal brings up three primary issues:
1.) Stores are not always square, or even rectangular. They have walls and extrusions which distort Wifi transmission and propagation.
2.) (Some, but not all) stores in a mall can be HUGE. Absolutely farther than LoS Wifi can carry, especially from a mobile device. The multitude of nodes and lack of indoor positioning systems can be an issue here.
3.) The effort of keeping a map up to date is directly proportional to its accuracy. Stores don't want location data, they want marketing data. From a participating store's perspective, what is more valuable: Having to spatially map their entire store (and keep said map up to date) to know that a customer stopped in the children's clothing department, girl's section, size 1-8 shoe rack
-- OR --
Only having to enter general descriptions of zones and knowing that someone stopped in the children's clothing department, girl's section?
>>51901254
OP said that the problem with this is that the stores don't share video feeds, which would throw a significant wrench in to this kind of system. Plus, ingesting video from hundreds of cameras could prove financially and computationally expensive. I have no knowledge of these domains, however.
>>51899975
You aren't good enough to make this system and fail spectacularly.
>>51901396
OP here. You are completely right, yes.
>>51901372
You are also right, I know it's legal.. I just want to know the opinion of fellow IT people.
>>51901399
Go fuck yourself.
>>51899975
1 step away from this
https://www.youtube.com/watch?v=7LcUOEP7Brc
Please don't be that guy.
>>51901528
I was thinking that this is amazing
>Firefox
Lwl
Wasn't there a company that got sued for doing this lately?
>>51901396
To clarify point 3,
It appears to me that the cost of having employees maintain the spatial index would outweigh the value of knowing precisely where people were.
Triangulation would provide somewhat more accurate data, which would require more processing power and more map fidelity. To demonstrate the difference in effort, allow me to pose a hypothetical:
A participating store closes down and entirely remodels the layout, and management needs to update the spatial map of the store.
For nearest-node, it's as simple as
1. Find the node and stand underneath it.
2. Describe the products nearby in the management interface.
For triangulation (or any other higher-accuracy solution), it becomes:
1. Walk through the entire store with this designated mapping device with two wifi adapters: One to continuously probe and feed the system location data, and one to actually communicate with the system over a real Wifi network.
2. Stop every 5 feet and note the products around you on to the machine.
For 30 nodes in a store, the first one could be accomplished in an hour, the other would require at least a full workday.
>>51901476
As someone concerned with privacy in the digital age, please add some sort of facility (A wifi network in a central point of the mall, named "$MALLNAME-opt_out"), which, upon connect, adds the MAC address to a system blacklist, which would filter incoming data from that MAC, and would purge the historical archives of matching data.
While I am dismayed by the advent of these opt-out systems, very few people would willingly opt-in, subverting the entire system's purpose. Further, if you don't build it, OP, someone else with less regard for privacy might.
You might also consider a one-way hash of the MAC as the unique identifier, as an added precaution.
>>51899975
They recently banned that here in the Netherlands.
>>51901626
>It appears to me that the cost of having employees maintain the spatial index would outweigh the value of knowing precisely where people were.
For individual stores.
But for big chains it's absolutely worth it.
They can use this to try out different concepts and accurately measure their effects.
Then copy the successful ones to other stores.
Knowing how many people walked by each product vs. how much was sold is very valuable information.
>>51899975
>is this morally wrong?
Yes.
That said, this is going to go ahead with or without you. At least if you're the one to do it, you can put some sort of safeguards in place to prevent tracking individuals, so that the only data they get out of the system has been anonymized to some extent, ie. just shows them overall trends.
>>51899975
Considering most of my posts have been walls of text, a question (buried in >>51901097) remains unanswered:
How exactly do the stores intend to get data out of this system? Will it need an API? Will it's output a map, or a search system? Will it simply show an aggregate of guessed demographics (25% estimated female shoppers in Store_A...)
>>51901695
This is true. Earlier in the thread I formed a large assumption: these are not large chains OP is dealing with.
I draw this conclusion from the fact that they're not only asking OP, an individual (not company) and a 4chan user, to build this system, when there are companies that actually produce these products. A large chain's corporate headquarters would not be asking an individual to create such a system, they would be taking quotes and evaluating systems from the industry (Cisco, Meru, etc.). This system would then be rolled out to their branches in a uniform manner. Single-branch management wouldn't likely have the authority to negotiate such a thing.
>>51901720
As I mentioned earlier, Apple has already begun using random MAC addresses for Wifi probes, which eliminates the issue of 1-1 MAC to Person identification for roughly half the mobile wifi userbase. Further, these patrons are already being recorded by CCTV, and, while I do agree that there *ought* to be a reasonable expectation of privacy in one's aggregate data (ALPR/ANPR location data, browsing history, etc.), none is currently recognised.
>>51899975
>I'm not sure about it to be honest. They also want me to make a system to figure out what type of people the customers are.. female.. male.. pregnant.. teen.. etc.
really not possible without requiring users to fill out a survey. desu you're better off getting customers to sign up for a rewards program tied to the [phone or an app, then using your location system to match locations to people
>>51900092
Church
>>51901899
Not true. There's the anecdote about the store (Target?) recognising pregnancy before the pregnant couple simply by analysis of purchasing habits. An amazing quantity of information can be gleaned from purchasing data, and browsing (literal store browsing, not Internet browsing) data can augment this further. But a survey might be a worthwhile addition to the system if there were a way to create a MAC to Respondent map and a way to motivate customers to fill it out.
>>51902039
raw purchase data still requires links to be made from demographic to purchases, so you're either using someone elses data (muh averages) or you're using your own by objectively linking a customer to his purchases, which requires you to get the data op is already looking for.
If I gave you a list of everything I bought in the last month, you would not be able to get much more than a vague understanding of who I am.
Here's how you do OP:
Go in saying you'll do it, get a few devices (paid for by them) and show a proof of concept. Make the program give you raw data dumps.
Once in place, say you'll do the stores on a per store fee based on area and once they're installed offer to do a monthly data report for them. If they say no, let them deal with the raw data dump. If they say yes, secretly build a program the auto compiles the report.
Easy cash, and for literally nothing.
>>51902101
I had two awful ideas when reading your post:
- If the mall's demographics support it (aka loads of teens), use social media to gain the needed data
- Have roaming employees record demographics of the patrons in zones
Neither of these address PURCHASE data, but as far as I can tell, OP isn't on the hook for purchase data.
>>51902166
We still don't know how OP's customers expect to get information out of the system at all, or in what format.
Questions for OP (if he ever comes back):
- (See above)
- How long do they expect you to provide updates for this system? Maintenance and hardware?
>>51900216
they going to do it anyways, get that money.
>sell the same product to others,
>then tell everyone what they are doing and how you regret help them
you'll be the hero with the money
[>and then nobody would care they got some personal data for profiling]
>>51902287
well they can get purchase data by matching location to transactions at the till.
i could see some sort of MITM attack to glean data sent to and from facebook to get ahold of a profile
>>51902287
forgot a question
- What is your stack? RPI OS? Server OS? Database? Programming Language?
>>51902291
And no business will ever trust him again.
>>51902331
OP's hardware isn't the Mall's network, its a second system entirely. He couldn't perform an MITM from his equipment. The wireless probes are infrequent (between thirty seconds and minutes if memory serves). For a quick purchase, OP's system might not even register them at checkout. The survey idea suggested earlier might be viable if you could tie it to a discount. This would uniquely identify the purchase and the MAC, linking the location track and the final purchase.
>>51902419
gotcha., i thought we was using the raspis to create a new distributed wifi network for the mall or whatever
How are you going to deal with directionality?
And how are you going to reduce the wifi power to make the sensors not step on one another?
I just made this last semester in my free time too track females on campus, duck morality
OP, look up Cisco CMX/Cisco MSE. Don't waste your time re-inventing the wheel.
>>51902469
space them far enough or attenuate them
How about one that uses cameras and face recognition instead of MAC address? Might also be able to tell age and gender. Is this possible yet?
>>51899975
In this world we live you gotta say fuck other people and get youra.
>>51902445
Facial recognition is not yet viable, plus (as has been stated a couple other times in the thread), the group of stores does not share CCTV feeds. Plus, that race+gender detection seems unreliable with general-purpose security cameras.
>mfw you will never do cool shit like OP
how do you even learn all of this
>>51899975
>>51899975
Look at it this way: if you hired Shaniqua to sit there and click one button every time Bill Smith walked into your privately owned property and another button every time Akhmed Abulazziz did, would it be morally incorrect?
Of course not.
But you should take it a step further: map it with your security system and now you'll have pictures of everyone. Map that to your (as yet undeveloped) facial recognition software, and present advertisements based on who that person is. Back to the Future anyone?
Honestly I'm surprised they aren't doing this already. But thanks for giving me yet another reason to never go to a mall.
>>51902528
Does such a system meet OP's budget in >>51900727? I expect Cisco shit would be in the $200,000+ range, but I'm not a network guy.
>>51902747
For me it was lots of autism and lots of free time: a potent combination. Despite my moral objections, this kind of stuff is cool as shit to me.
>>51902747
Just stackoverflow your way through every problem.
>>51902878
Those guys quite often explain questions they set themselves, not what OP asks, and those questions quite often don't have anything to do with OPs quesition
you wont get much out of that. if you capture the probes coming from the device and spoof the ssid to get it to connect, that can easily be considered unauthorized access to the device and could get you in trouble.
otherwise you are looking for probes and logging them, which will work but you are talking lots of access points to actually get accurate locations.
if the device connects to the wireless network it is the same as before, you need a lot of access points to make it worthwhile.
in reality its not that good of idea to put into practice. something like rfic/nfc is a far better solution, but again in practice the odds of actually getting people to use it is slim to none.
its a good idea on paper and not really 'morally wrong', just hard to actually implement to get any real results
i work in security and i do wireless stuff, even IF you have the money to dump access points everywhere, getting information like women/man/pregnent is going to be kind of hard to substantiate
>>51902933
Say OP gets patrons to fill out a survey in exchange for a universal 10% discount. Then, he gets the demographics data (survey results), unique ID (MAC used to fill out the survey), location data (passively-collected via his RPi network), and purchase data (when discount code is claimed, the purchased items are logged to the system).
If you throw in some fancy statistics for inferences and a nice UI for the stores' management you've got a hell of a product on your hands.
>>51903257
again, good on paper but in practice not much. it would be tough to actually get a real response from having people need to use their phone to fill out an account.
even actually trying to track someone with accuracy in a closed quarters like that would be a nightmare, and to actually have that many access points together you would need expensive shit with proper wireless controllers. you are talking a lot of money just to get the network working and you STILL have to put those AP's basically at eye level
>>51899975
Already exists, we use it in home depot
>>51899975
Just put a chip in the baskets/carts, doubles as an alarm if the person tries to take them off property.
>>51903493
>Already exists
Yeah, but OP has been contacted to build it. If someone comes to you and says they'll give you $50,000 to reinvent the wheel, why would you say no?
>>51903539
because its a waste of time when you can simply purchase a system already built
>>51903584
That advice is rarely considered on /g/.
>Basically, they want me to place small wifi sensors everywhere and using the MAC addresses, they want me to make a mapping of how most customers walk through their stores.
sounds totally reasonable. no private information being mined so i'm not seeing the issue he-
>They also want me to make a system to figure out what type of people the customers are.. female.. male.. pregnant.. teen.. etc.
lel. while i can't think of any way that would work, that would indeed be pretty crappy. honestly how would you even do that though? i'd ignore the request
>>51899975
If you can trace the MAC addresses to individual customers, then this could be a privacy disaster. Think about it: if the customers found out they were being identified, and tracked around the store, wouldn't they be a bit concerned? If I were you, I'd look up the privacy laws and see whether this is legal.
>>51903750
you don't have to design the system to do that. it would be rather trivial to design the system to prevent exactly that
>>51899975
wait till you meet my mighty morphing MAC address.
>>51903815
wait till u meet my dick fag!!!
You're setting up a botnet
>>51903750
>Think about it: if the customers found out they were being identified, and tracked around the store, wouldn't they be a bit concerned?
>what is CCTV
>>51899975
Are you the only one they approached?
