The 'deanonymization' attack uses social network groups as well as some traditional browser history-stealing tactics to narrow down and find the user behind the browser. The researchers were able to deanonymize more than half of the users in their initial test using their attack method, which entailed their joining and crawling groups within social networks, such as Germany's Xing business social network and Facebook, using a fake profile. They then matched pilfered browsing histories with social-network group members to "fingerprint" and identify them.
"Without using the group info, an attack that only uses history stealing is infeasible in a real-world scenario. So, in fact, it is the combination of history-stealing and group information that is novel," says Gilbert Wondracek, a post-doctoral researcher with the International Secure Systems Lab of the Vienna University of Technology in Austria, who co-developed the proof-of-concept.
Criminals could use this for phishing and targeted attacks. The attack requires only that the victim visit a malicious Website that contains the attack code -- there's no malicious link, per se. "We could put the attack code on a Website that contains a political, dating, religious, [or other] forum. If someone posts anonymously to this Website, there is a chance that we could find out the social network profile for this person," Wondracek says. "Since social network profiles contain a wealth of info and, per definition, the friends of this person, blackmailing is also an option."
Wondracek says he and fellow researcher Thorsten Holz had wondered how the well-known history-stealing technique could used to unmask online users via their social networking profiles. History stealing allowed them to peek at a user's URL browsing history to see if he had visited specific social network groups -- sports-related or other groups that friend or fan organizations, for instance -- that the researchers had joined.
"We can now perform an intersection and find out that there are just a few people in the whole social network that belong to exactly these ... groups. The group fingerprint is rather unique among all users," Wondracek says.
Then the attacker uses history-stealing once again to check for links that are similar to each member of the groups.
The researchers say that while their PoC was for Xing, it can work with any other social network. They crawled 7,000 public groups in Xing and found around 1.8 million users belong to at least one group. "These users are vulnerable to our attack," Holz blogged recently.
Volunteers from Xing can participate in the experiment via the researchers' demo Website here. The more regularly a Xing user participates in groups on the social network, the more likely he will be deanonymized by the PoC.
There is no fix for this attack, but workarounds include turning off browsing history or using private-browsing mode. Wondracek says the only protection social networks could provide is to change the way their Web applications use hyperlinks to move information from one point of their site to another in "keep state." Xing has implemented this as part of its response to the attack research, he says.
"I was -- and am still -- quite surprised that, a, getting the group data was so easy, and, b, almost all social networks use URLs that leak private information," Wondracek says."The attitude behind this is pretty scary from our maybe naive point of view."
Have a comment on this story? Please click "Discuss" below. If you'd like to contact Dark Reading's editors directly, send us a message.