It seems highly likely to me that the tracking data retained by iPhones and (currently) backed up in the collections.db database is fairly innocuous.
I pulled together some Perl scripts to analyse the data and have come to the following conclusions:
- The CellLocation Table (8000 entries) is a subset of Apple’s global cell tower location database that is cached locally. The timestamps in this only makes sense if they are part of a batch update from some other source. In my case there about 100 distinct groups of location and they don’t match where I’ve been at any level smaller than about 50km and on a timescale of days. There are some really odd entries too – one in northern Italy that I haven’t even flown within 500 km of in the past year.
- The WiFi Location table (60000 entries) has similar characteristics. It’s got groups of AP locations that also make no sense. My table correctly has a bunch of locations in Bracknell during March but there are also more than a hundred entries much further North at the same time in a curious east <-> west line that only makes sense if there is some caching process that pre-fetches a bunch of these to improve non-GPS positioning performance. All the MAC addresses listed appear to be infrastructure (AP’s) rather than user systems so the general threat from it is pretty limited.
- The CellLocationLocal Table looks much more like tracking data but it only has 106 entries. All of these have unique timestamps, include altitude, speed and heading information and most importantly all seem to correspond very accurately (to within a few metres) with times and places that make sense even though they claim the accuracy is only 1500m. Thinking carefully about it all of these seem to correspond with times when I had the GPS function enabled. It looks to me as if this is data that Apple might actually be likely to harvest and could be sending home in order to improve their Cell Phone location maps.
Having discussed this with some folks and after reading the online commentary I think it’s pretty safe to say that this looks like a storm in a teacup. There is certainly some tracking data in the file, but it’s privacy risks are a lot less than they seemed. Apple don’t appear to be gathering location data that includes the DeviceID or any other details that could uniquely identify the user or phone. They could be sending that back along with such an identifier but I doubt that they would – having these tables structured in this way only makes sense if the primary purpose is to enable a local cache for quick location lookups.
Ultimately I’m a bit disappointed – I’d wanted to have a reliable record of where I’d been over the last year – but on the plus side it really doesn’t look as if Apple are harvesting data that could be used to spy on me. We’ll have to se what comes out on this over the next few days but I’m not losing any sleep over this.