ABSTRACT
This chapter addresses the problem of detecting anomalous entities in dynamic social networks by considering dynamic infinite features of network entities and dynamic feature cascading. To cope with this problem a statistical approach for modeling of dynamic social networks is developed, that is capable of modeling the cascade behavior of infinite features, also referred to as social influence. The proposed approach can identify the birth, death, and longevity of individual features as well as networks entities taking and giving up the features in order to model dynamics of network structure. The chapter shows that the relationships between entities with different features are also considered in addition to the ones between entities with the same features. It describes how to detect anomalous entities, including anomalous links and anomalous nodes in dynamic social networks. Anomaly detection in complex networks is a vibrant research area and this problem has been addressed in many literatures.
