Detecting Anomalous Behaviors Using Structural Properties of Social Networks

Abstract

Abstract. In this paper we discuss the analysis of mobile networks communication patterns in the presence of some anomalous "real world event". We argue that given limited analysis resources (namely, limited number of network edges we can analyze), it is best to select edges that are located around 'hubs' in the network, resulting in an improved ability to detect such events. We demonstrate this method using a dataset containing the call log data of 3 years from a major mobile carrier in a developed European nation.

Parallel Experiment for Urban Rail Emergency Evacuation: An Approach for Hub Identification

“…We can reliably approximate the weights w ij by a power-law dependence 9-12 : w ij = (k i k j ) q , where q is a weight parameter that should be calibrated depending on the real data of the concerned networks, and k i and k j are the degrees of stations i and j, respectively. Empirical evidence supports this definition, [9][10][11][12] and researchers widely use it on real weighted networks.…”

“…Furthermore, using the identified hubs, operators can detect some anomalous events or emergencies in the network even if they can't observe all of the nodes. 12 both the station's internal passenger flow capacity and the importance of its external neighbors. Figure 3 shows that an equal number of passengers in an emergency node can evacuate more quickly to neighbor 1 than neighbor 2.…”

Image-oriented economic perspective on user behavior in multimedia social forums: An analysis on supply, consumption, and saliency

“…Existing work on saliency detection for social multimedia services includes Heard et al (2010) and Altshuler et al (2013). They formulate the problem as that of detecting network nodes which exhibit abnormality in terms of structural patterns and connectivity.…”

Personalized Economy of Images in Social Forums: An Analysis on Supply, Consumption, and Saliency

“…Existing work on saliency detection on social multimedia services mostly formulates the problem as detecting nodes that exhibit abnormal structural patterns such as connectivity [13], [14]. Rather than focusing on structural properties, our saliency detection approach focuses on detecting salient users based on the actual visual content they supply and consume.…”

“…In particular, our work focuses on identifying salient 'suppliers' and images posted by them, because supply is arguably a more important contributing factor towards the success and failure of social multimedia services, more so than consumption. A novel aspect of our approach is that a user's saliency is measured based on the visual content they supply, i.e., their supply profile, while most existing work [13], [14] address this problem by analyzing abnormal structural patterns in social networks such as properties of friendship connectivity. In addition, we also show that users' saliency may be time-varying -a user who exhibits regular behavior in general may still show salient patterns during certain time intervals, and by detecting such intervals, we can identify key images that contribute significantly to increase the user's saliency.…”