Dynamic Identity Classification on Real-Time Randomized Social Networks

Social network analysis is a critical topic in the world of data mining and active security and privacy threats. In particular, the capabilities of social networks to undermine a user's privacy or information security is of major concern. A user's information can often contain sensitive or identifying content, thereby creating a threat of abuse if inadvertently disclosed or obtained. Through social network analysis, one can obtain an association between a user's behaviors and identity. Such an association compromises the privacy of users in a social network, particularly in real-time, hence the need arises for a remedy. Without a real-time solution, the security and privacy of users becomes delegated to a single point of failure in access control (AC) mechanisms. If the AC is subverted, vulnerabilities in social networks can be exploited to obtain the user's identity. This paper utilizes Artificial Neural Networks (ANN), in the context of dynamic social networks, to identify actors. Obtained results demonstrate a high average user identification rate on an arbitrary subset of the network.