Schneier on Security
A blog covering security and security technology.
« Terrorist Havens |
| Hacking Two-Factor Authentication »
September 21, 2009
Inferring Friendship from Location Data
For nine months, Eagle's team recorded data from the phones of 94 students and staff at MIT. By using blue-tooth technology and phone masts, they could monitor the movements of the participants, as well as their phone calls. Their main goal with this preliminary study was to compare data collected from the phones with subjective self-report data collected through traditional survey methodology.
The participants were asked to estimate their average spatial proximity to the other participants, whether they were close friends, and to indicate how satisfied they were at work.
Some intriguing findings emerged. For example, the researchers could predict with around 95 per cent accuracy who was friends with whom by looking at how much time participants spent with each other during key periods, such as Saturday nights.
According to the abstract:
Data collected from mobile phones have the potential to provide insight into the relational dynamics of individuals. This paper compares observational data from mobile phones with standard self-report survey data. We find that the information from these two data sources is overlapping but distinct. For example, self-reports of physical proximity deviate from mobile phone records depending on the recency and salience of the interactions. We also demonstrate that it is possible to accurately infer 95% of friendships based on the observational data alone, where friend dyads demonstrate distinctive temporal and spatial patterns in their physical proximity and calling patterns. These behavioral patterns, in turn, allow the prediction of individual-level outcomes such as job satisfaction.
We all leave data shadows everywhere we go, and maintaining privacy is very hard. Here's the EFF writing about locational privacy.
EDITED TO ADD (10/12): More information.
Posted on September 21, 2009 at 1:41 PM
• 14 Comments
To receive these entries once a month by e-mail, sign up for the Crypto-Gram Newsletter.
"researchers could predict with around 95 per cent accuracy who was friends with whom by looking at how much time participants spent with each other during key periods, such as Saturday nights."
Having spent the last three weeks in hospital I dread to think what these researchers might infer from my mobile phone location...
The US Gov a few years ago decided that they nolonger needed expensive on the ground unreliable "Human Assets" so cut back on HumInt and went head over heals with technology such as SigInt and ElInt and overhead observation.
Belatedly the US Gov has found out that hard data from TechInt sources whilst acurate does not provide the richness of information that HumInt does...
Time/Location data of this form is TechInt, it provides little or no information from which future actions / associations can be infered.
Importantly it only tells those looking at the data where a phone is not where it's nominal owner is.
Some people already regard this sort of information as being usefull to them for covering up their activities and routienly swap their mobile phones around.
This is not just at the criminal but business levels I suspect that those who can think about such issues and may be effected by them are already taking steps to look like "profile A" not "profile X"...
I am surprised this study is recent, as I can remember reading, a few years back, of the same kind of study. (In IEEE Security & Privacy, if memory serves me corectly)
In fact, that previous study had claimed to have "accidently" discovered a secret relationship in the staff taking part of the experiment.
So it kind of smells like "déjà vu"
Schneier.com is a personal website. Opinions expressed are not necessarily those of Co3 Systems, Inc.