Protecting Private Information in Online Social Networks

Because personal information can be inferred from associations with friends, privacy becomes increasingly important as online social network services gain more popularity. Our recent study showed that the causal relations among friends in social networks can be modeled by a Bayesian network, and personal attribute values can be inferred with high accuracy from close friends in the social network. Based on these insights, we propose schemes to protect private information by selectively hiding or falsifying information based on the characteristics of the social network. Both simulation results and analytical studies reveal that selective alterations of the social network (relations and/or attribute values) according to our proposed protection rule are much more effective than random alterations.