Temasek Defence Systems Institute (TDSI) would like to invite you to attend the research presentations (Free Admission for All). 

Date: 7 August 2017 (Mon)

Time: 3.30-5pm

Venue: NUS, Faculty of Engineering, Block E2 Level 3 Room 3 (E2-03-03), Singapore 117579

Speaker:A nalyzing the Robustness of Centrality Measures

LTC Jon Roginski

Director, Network Science Center

Assistant Professor

United States Military Academy, West Point, NY


Centrality measures are often used to identify the “importance” of a vertex in a graph from a certain perspective. Degree centrality highlights a measure of popularity by identifying those vertices with the most adjancencies. Eigenvector centrality shows vertices that are important because its neighbors are important. Closeness centrality highlights those vertices that indicate travelling “shortcuts” throughout the network. Betweenness centrality shows gatekeepers of graph communities. The newly-identified distance centrality provides a measure of a vertex’s impact on topological structures. Each of these measures has its place, depending on the desired insight to be gained or decision to be informed. This work provides an analysis of each centrality’s behavior under imperfect information. Does one centrality or another exhibit robustness under uncertainty or fragility? Which centrality measure is the most reliable in situations that we know the network information is imperfect or wrong? Answering such questions is critical to inform decision makers in situations where taking the wrong action at the wrong time may be harmful to the organization’s operating efficacy.

An Analysis of Terrorist Networks


Associate Professor Ralucca Gera

Department of Applied Mathematics

and Center for Cyber Warfare

Naval Postgraduate School, Monterey, CA


Dark networks, modeling illegal or covert activities, are of great interest to intelligence analysts in understanding their behavior.  As data about these networks is not easily obtainable, we are interested in effectively modeling social covert networks as multilayered networks.  We present a network science analysis of these models, particularly identify communities and search for people of interest in these social networks.