The power of coincidence may help identify the social ties of even the most privacy-conscious web user, a new study of image-sharing site Flickr suggests.
Last year David Crandall, now at Indiana University in Bloomington, analysed the geotags on Flickr photographs – the latitude-longitude coordinates that most smartphones and many cameras add automatically to photographs. Given a large enough photograph collection, he showed that it was possible to create accurate global and city maps.
Crandall and his colleagues have now shown that the same information provides an online record of the coincidences in which friends often find themselves in roughly the same place at the same time.
To spot these coincidences, the team used 38 million geotagged and time-stamped photographs from Flickr. They arbitrarily divided the planet into cells of 1 degree of latitude by 1 degree of longitude – squares roughly 80 kilometres to a side at middle latitudes, although much larger near the equator. Then they compared the time and location stamps of uploaded Flickr images with the photographer's declared friendships in the website's social networking section.
Crandall's team found that two people have a 5 per cent chance of knowing each other if on three separate occasions they take an image within the same cell on the same day. This is 300 times greater than the 0.0134 per cent chance of two randomly chosen Flickr users knowing each other, they say.
The likelihood jumps to nearly 60 per cent, or 5000 times the random probability, for two people who take images in the same cell on the same day on five occasions. What's more, these numbers are likely to be an underestimate, Crandall says, as many Flickr users keep their contact lists private or do not use the website's social networking feature.
Crandall admits that, on the other hand, the data might be skewed to favour such an outcome because people may be more likely to take photographs when out with their friends. But he points out that this effect is counterbalanced by the fact that people also tend to upload photos taken at big events like concerts, which are attended by thousands of strangers who will do the same thing. Quantifying such factors and controlling for them remains a challenge, he says.
The research could have privacy implications, as it shows that even a collection of random photographs of landmarks that feature no people at all can send strong signals about your social ties.
"We really don't understand the privacy implications of the data about us online," says Crandall. "Even if people knew that their geopositions were being made public, I don't think they would know that can reveal who their friends are."
This post originally appeared on New Scientist.