Author Topic: Accuracy of IP geolocation?  (Read 622 times)

georgem

  • Tech Admin
  • ******
  • Posts: 842
    • View Profile
Accuracy of IP geolocation?
« on: March 02, 2016, 09:02:43 pm »


Quoting coins101 Big Data Overview Paper: https://docs.google.com/presentation/d/1qyExM5c1Ws7eUEj3J_gJ76zFe2WQaLdqJHMbJIJ_jng/edit?pref=2&pli=1#slide=id.g81d91d174_1_11

Code: [Select]
"IP data can be converted to non-personal Geo reference data before being released to third parties."
How acurate are IP to geolocation conversions anyway?
What are their limiting factors?
Can it be improved?
What margin of error are we talking about here?

Please discuss this very important topic in this thread.

rhinomonkey

  • Jr. Member
  • **
  • Posts: 64
    • View Profile
Re: Accuracy of IP geolocation?
« Reply #1 on: March 10, 2016, 06:52:19 am »
I went back and found some old, though recent-ish posts about this particular area of Big Data research.

Here are some of the posts/ conversations that took place amongst Georgem and Coins101:

Coins101:

"Some context:

My primary interest is IP geolocation data related to mobile SPVs. That's where I see the really big value in Big data.

In theory, we should be able to deploy 'let this app use my location' which is accurate to within a few meters.

User money spending habits + location = holy grail for, erm, everyone.
"

Georgem

Quote from: georgem
Quote from: coins101
Some context:

My primary interest is IP geolocation data related to mobile SPVs. That's where I see the really big value in Big data.

In theory, we should be able to deploy 'let this app use my location' which is accurate to within a few meters.

User money spending habits + location = holy grail for, erm, everyone.

Yes, I'm aware of that. Including SPV nodes increases the reach of the "scanned area".

But then, as a precaution, should we even communicate to the outside world (big data market) if a user was sitting behind a mobile SPV or a desktop pc's full node or whatever other means of access he used?

We shouldn't.

But back to geolocation: is this really accurate within a few meters? Maybe in some urban areas, but this can fluctuate wildly.

Coins101:

Quote from: coins101
Quote from: georgem
...

We shouldn't.
...

Ah, the authoritarian. You can choose any colour you like, as long as its the one I like approach  ;D

Let people choose to opt in (help bitcoin), or opt out (this stuff creeps me out).

IP addresses are poor for locating people to within a few miles.

Mobile SPVs have a far better gps style location system this is accurate to within a few meters, for those that hit that opt-in button.

edit:

So you can choose

> I'm buying drugs today and then visiting a stripper. I'll opt out.

> I'm buying beans today, who gives a shit who knows that. I'll opt-in and help Bitcoin

Georgem:

Quote from: georgem
Quote from: coins101
Quote from: georgem
...

We shouldn't.
...

Ah, the authoritarian. You can choose any colour you like, as long as its the one I like approach  ;D

Let people choose to opt in (help bitcoin), or opt out (this stuff creeps me out).

IP addresses are poor for locating people to within a few miles.

Mobile SPVs have a far better gps style location system this is accurate to within a few meters, for those that hit that opt-in button.

edit:

So you can choose

> I'm buying drugs today and then visiting a stripper. I'll opt out.

> I'm buying beans today, who gives a shit who knows that. I'll opt-in and help Bitcoin

I'll defend privacy even if I have to become an authoritarian, lol! hitler himself!

Is there any current SPV wallet out there (for any coin) that is asking the user for access to its mobile's geolocation data?

I don't think so, for the same reason my bank doesn't ask me about geolocation every time I log into e-finance.

Since SPV wallets are inherently insecure devices, is it clever to broadcast the very position of such a device to a network?

I'm just asking.... it sounds like something that will be very tough to sell to the masses. ("... and that's supposed to help bitcoin??" they will ask, ... "by compromising my security?")

So I guess we'll be stuck with IPs.

Your thoughts? Maybe I'm missing something.

Coins101:

Quote from: coins101
Quote from: georgem
....
Your thoughts?

Don't get hung up on data = a specific person.

Put the biggest filter you can find into the system, so data doesn't equal a specific person.  Use zero knowledge proofs so that all you can tell is crypto transaction, $15, at x place and at y time. That's the most valuable part of the transaction.

Holding personal data is a headache and generates regulatory issues.  Don't want it. Don't want it. Don't want it.

I've said it many times before. Personal data: Don't want it.

People that buy this type of data don't care about tracking down individuals to sell them a can of beans. 

Roger Ver should be walking into Starbucks, right about now...quick, go there now with large pack of washing tablets and a 15% coupon....

Of the $1bn spent in Star City last month; 10% was in that area; 5% was over there outside that mall; 30% was over there between the hours of 8pm and 3am; etc.

Right, lets get our marketing strategy right so we can advertise where most people buy xyz; lets make sure we're stocked up with xyz during 8pm to 3am; etc.

See, anonymous data can be really useful and valuable.

edit

And.....it is very on point:  http://www.comsoc.org/netmag/cfp/mobile-big-data


Georgem:

Quote from: georgem
Quote from: coins101

Put the biggest filter you can find into the system.

The filter (or shall we say obfuscator) would have to be on the mobile device itself.
But even then, this doesn't stop the weird feeling users are going to have:

"why is this wallet asking for access to my geolocation?"

"Ah ok, my location is safe, and I'm supposedly helping bitcoin miners by allowing spreadcoin to sell my data to merchants."

"But what do I have to win from this?"

I predict that no one is going to want to participate in this scheme, just to help bitcoin miners.

People can always choose to help miners directly.

The mere act of buying bitcoin regularly is what helps miners the most. Obviously.


People are going to think that this "big data" scheme is like a (be it voluntary) taxation charity scheme,
where a middleman (spreadcoin, not related to bitcoin miners) is selling something to merchants (again, not related to bitcoin miners)
so that finally some of the profit made along the way is being shared with the poor miners.

Doesn't sit right with me, if I heard it explained to me this way.

Quote from: coins101
Holding personal data is a headache and generates regulatory issues.  Don't want it. Don't want it. Don't want it.
I've said it many times before. Personal data: Don't want it.

That's the spirit!

Quote from: coins101
See, anonymous data can be really useful and valuable.

Yes, that's why I still entertain the idea of decentralized big data, because I think it actually can be done in an anonymous way.
There is a way, we just have to find it.


These convesations took place on pg 164 - 165 on BTCT. Link: https://bitcointalk.org/index.php?topic=1045373.3260

georgem

  • Tech Admin
  • ******
  • Posts: 842
    • View Profile
Re: Accuracy of IP geolocation?
« Reply #2 on: March 10, 2016, 03:40:05 pm »
Thanks.

So as it seems, for big data, IP geolocation doesn't have much use because it is very inaccurate.
(But it depends what goals you have planned for big data.)

So, coins101 is thinking about ways of how his big data methodology can get its hands on actual GPS data of mobile devices.

Currently I don't think there is any mobile wallet out there that actually even DARES to ask users for permission to access their GPS data, because from a security standpoint it's very problematic.
And people will instantly get turned away by that.

So I guess we are stuck with IP geolocation, which ... thru its inaccuracy adds a certain level of obscurity that is desirable from a privacy standpoint.

In an early big-data experiment, we could try and just figure out ways of analyzing traffic based on countries alone.
Suppose we offer information about which country has how much percentage of certain altcoin (or bitcoin) traffic,
that alone would be valuable for investors who try to establish a business in a certain country and need to know how much of the global usage of a certain altcoin actually falls within the same geographic location.

That's how I would start, and build from there.