Complex network tools to enable identification of a criminal community

Magalingam, P 2015, Complex network tools to enable identification of a criminal community, Doctor of Philosophy (PhD), Mathematical and Geospatial Sciences, RMIT University.


Document type: Thesis
Collection: Theses

Attached Files
Name Description MIMEType Size
Magalingam.pdf Thesis application/pdf 4.71MB
Title Complex network tools to enable identification of a criminal community
Author(s) Magalingam, P
Year 2015
Abstract Retrieving criminal ties and mining evidence from an organised crime incident, for example money laundering, has been a difficult task for crime investigators due to the involvement of different groups of people and their complex relationships. Extracting the criminal association from enormous amount of raw data and representing them explicitly is tedious and time consuming. A study of the complex networks literature reveals that graph-based detection methods have not, as yet, been used for money laundering detection. In this research, I explore the use of complex network analysis to identify the money laundering criminals’ communication associations, that is, the important people who communicate between known criminals and the reliance of the known criminals on the other individuals in a communication path.

For this purpose, I use the publicly available Enron email database that happens to contain the communications of 10 criminals who were convicted of a money laundering crime. I show that my new shortest paths network search algorithm (SPNSA) combining shortest paths and network centrality measures is better able to isolate and identify criminals’ connections when compared with existing community detection algorithms and k-neighbourhood detection. The SPNSA is validated using three different investigative scenarios and in each scenario, the criminal network graphs formed are small and sparse hence suitable for further investigation.

My research starts with isolating emails with ‘BCC’ recipients with a minimum of two recipients bcc-ed. ‘BCC’ recipients are inherently secretive and the email connections imply a trust relationship between sender and ‘BCC’ recipients. There are no studies on the usage of only those emails that have ‘BCC’ recipients to form a trust network, which leads me to analyse the ‘BCC’ email group separately. SPNSA is able to identify the group of criminals and their active intermediaries in this ‘BCC’ trust network. Corroborating this information with published information about the crimes that led to the collapse of Enron yields the discovery of persons of interest that were hidden between criminals, and could have contributed to the money laundering activity. For validation, larger email datasets that comprise of all ‘BCC’ and ‘TO/CC’ email transactions are used. On comparison with existing community detection algorithms, SPNSA is found to perform much better with regards to isolating the sub-networks that contain criminals.

I have adapted the betweenness centrality measure to develop a reliance measure. This measure calculates the reliance of a criminal on an intermediate node and ranks the importance level of each intermediate node based on this reliability value. Both SPNSA and the reliance measure could be used as primary investigation tools to investigate connections between criminals in a complex network.
Degree Doctor of Philosophy (PhD)
Institution RMIT University
School, Department or Centre Mathematical and Geospatial Sciences
Subjects Applied Discrete Mathematics
Keyword(s) Criminal network
Trust
Shortest path
Leave-one-out
Suspect
Investigation
Betweenness
Intermediate node
Reliance
Versions
Version Filter Type
Access Statistics: 230 Abstract Views, 969 File Downloads  -  Detailed Statistics
Created: Tue, 13 Oct 2015, 13:05:07 EST by Keely Chapman
© 2014 RMIT Research Repository • Powered by Fez SoftwareContact us