Scientists have developed a new generic algorithm based on machine-learning to detect fake accounts on social network platforms including Facebook and Twitter, an advance with considerable potential for applications in the cyber-security arena. "With recent disturbing news about failures to safeguard user privacy, and targeted use of social media to influence elections, rooting out fake users has never been of greater importance," said lead researcher Dima Kagan from the Ben-Gurion University (BGU) in Israel.
The study showed that the algorithm is generic, and efficient both in revealing fake users and in disclosing the influential people in social networks. "Overall, the results demonstrated that in a real-life friendship scenario we can detect people who have the strongest friendship ties as well as malicious users, even on Twitter," the researchers said.
Based on machine-learning algorithms, the new method, detailed in the journal Social Network Analysis and Mining, works on the assumption that fake accounts tend to establish improbable links to other users in the networks. It constructs a link prediction classifier that can estimate, with high accuracy, the probability of a link existing between two users. It also generates a new set of meta-features based on the features created by the link prediction classifier.
Using the meta-features, the researchers, constructed a generic classifier that can detect fake profiles in a variety of online social networks. "We tested our algorithm on simulated and real-world data sets on 10 different social networks and it performed well on both," Kagan said. Previously, researchers from the BGU had developed the Social Privacy Protector (SPP) to help users evaluate their friends list in seconds to identify which have few or no mutual links and might be "fake" profiles.