Covert Network Analysis to Detect Key Players using Correlation and Social Network Analysis

Abstract

A sudden escalation of terrorist events entices many researchers to divert their attentions to counter-terrorism field, and contribute in developing new techniques and methods for analysis, identification, exposing and prediction of terrorist events by using latest technologies. To be more effective, terrorist keep their identity secret in hierarchy of any organization to escape from the eyes of law and enforcement agencies. But to accomplish the fatal activities of terrorism successfully, they must need to communicate with others in network to make plans. The stealthy hierarchy of such terrorist organizations could be exposed by the pattern of collaborations and communication among members. The social network could be defined as “A social collection made up of social actors like persons or organizations and a compound set of links between these actors”. This definition inspires us to treat terrorist networks as the normal social networks, so that different social network analysis techniques could be applied on such networks to extract their hierarchy and useful information, which could help us to predict something about terrorist and their activities. The prediction about stealthy hierarchy of terrorist networks would expose the significance of every node in the network. This could help us to predict something useful in order to destabilize the network, which eventually results in immobilization of terrorist to accomplish their evil intents. From our analysis, we come to know that the traditional measures for social network analysis are not capable or related to above mentioned problem, except from the one the relative degree. The inability of traditional measures is caused by the stealthy hierarchy and also hidden intents of the terrorist. A node in terrorist network might not be prominent but that particular node might be the leader of the network. Removing that node from the network will help in easily destabilizing the network. Current methods of social network analysis mainly focus on the different degree measures considering nodes and edges of the network, while there is still a lot of work need to be done keeping the focus on the communication between nodes. These aspects of social networks like secrecy and lack of analysis to detect importance of node in network using nodes communication inspires us to propose a framework to detect key players in covert network using text mining. This thesis, describes to build a model to find the correlation between a data dictionary and communication of nodes, to evaluate and detect the key players having the highest similarity with data dictionary, which consists of words or terms mostly used by terrorist for their organizational structure, terrorist activities, planning etc. Our research work mainly presents this novel model and analysis to detect key player using this model. In this thesis, we used Enron email dataset to test and validate our novel proposed model. In this thesis, we also develop a data dictionary from the selected dataset, as a pre-defined data dictionary was not available. We also devise a preprocessing module, which could be used in any text many application. Till date according to our understanding no such model exist, which is dealing with the conversation data of nodes for key player detection.