Where were you on the night of ... ?

With the move by twitter to license its history of tweets, an interesting and heated debate intensifies about the moral and legal aspects of the use of fine grained location intelligence information acquired from voluntary contributors who mostly have no idea what they have given away.

Twitter has had a geolocation option since the 19th November 2009. Of course you have to turn it on in the settings page.

Given the lovely juicy and clever location based applications that have used the Twitter Geo API there is reason to suspect that a large number of tweets have location information attached. Challenge: If you are on twitter, do you know whether your location flag is turned on?

Even more interesting is that the twitter API has reverse geocoding. The location information is related to a physical entity rather than coordinates and so the association of the tweet data is immediately pertinent.

Rudimentary web research suggests there are 290 million tweets a day.

The Location Forum has set itself the task of looking at the implications of location data privacy. An assertion by the group is that location information in data, leads to inferences that may not be immediately obvious and penetrate privacy barriers in manners that are poorly understood and certainly not understood by the casual user of social media (like tweeters). This is referred to as "the power of where".

The Location Forum is working on a study to attempt to translate this into useful guidelines but in the meantime, in keeping with the title of this blog, lets look at a few examples of the dark side.

1.Hesitant about office romances, Jim quietly offers Julie tickets to see a movie. Jim tweets in the lobby that Hugo looks like a great movie and Julie tweets that the popcorn in the theatre is expensive. Busted ! But did either of the protagonists have any idea how effectively they were giving each other away ?

2.Facebook based Burglary

3.Intent on achieving a majority government, a political party charters a company to find ways to be sure of a majority. A study of multiple census variables establishes projected behaviour patterns over census areas. The census areas are of course generalized so that nobody can identify individuals. However, the company also accesses survey data available for purchase that is generalized over a different geometry set. This data is added to with behaviour data (purchased again) obfuscated over another geometry fabric. Combining and analysing the data infers knowledge about fractions of geometries that was otherwise impossible to reverse engineer. The data inferred is then matched to actual addresses (again available for purchase). A reverse lookup of the phone numbers (purchased) is followed by bot calls all those phone numbers to confuse the voter about which voting station to go to (from freely available data).

Perhaps this last one has also happened already.