Walter Scott, Jr. College of Engineering

Graduate Exam Abstract

Shashika Muramudalige
Ph.D. Preliminary
Nov 15, 2019, 9:00 am - 11:00 am
Physics Conference Room
Abstract: Identifying and analyzing latent and
emergent patterns in behavioral data
is significant in domains such as
homeland security, consumer
analytics, cybersecurity, and
behavioral health. We propose to
develop algorithms, tools, and
techniques for the identification of
such behavioral patterns with focus on
domestic radicalization. The outcomes
of the proposed research will be useful
for administrative bodies and law
enforcement authorities to counteract
the threat of homegrown violent
extremism, which has become a vital
issue not only in the west but also in
many parts of the world. The severity
and frequency of these attacks have
significantly increased in recent years,
resulting in significant loss of lives and
undermining of democratic institutions.
Identifying home-grown violent
extremists and preventing future
attacks are exceptionally challenging
due to the lack of behavioral trajectory
models during radicalization. In many
cases, preparatory tasks or even
attacks are committed as groups and
detection of such collective activities
complicates the investigation because
individual indicators can be innocuous,
but their significance is only exposed
by analyzing the integrated behavior of
the group. The tools developed in this
research will enable social scientists
and others to deal with larger volumes
of data with automated analytic
capabilities. Our research also aims at
overcoming limited data availability for
domestic radicalization and preserving
the privacy of individuals by generating
synthetic datasets that mimic
characteristics of real data.

In our study, we propose an end-to-
end investigative pattern detection
framework to identify specific
behavioral patterns of individuals and
groups using machine learning and
graph pattern matching techniques.
The proposed framework is applicable
to many investigative domains with
adequate datasets. Our work will
facilitate large-scale mining of detailed
forensic biographies of known
American jihadists, which are
extensively studied by social scientists.
Algorithms for emergent and latent
patterns will be developed and
validated. We introduce diverse
techniques and approaches for
different components of the proposed
framework; behavioral indicators
extraction from text sources,
investigative graph search in social
networks, and synthetic profile
generation. We propose to integrate
different NLP approaches to identify
radicalization behavioral indicators in
text documents. An investigative graph
search to detect potential risk profiles
and groups in large social networks
will be presented, with scalability
achieved by leveraging appropriate
data storage mechanisms. A novel
synthetic profile generation technique
will be introduced while preserving the
statistical characteristics of the actual
data. We will also address the
scalability, robustness, and efficiency
of the proposed techniques to affirm
the applicability to real-world scenarios
that require mining and processing of
billions of records.
Adviser: Anura Jayasumana
Co-Adviser: N/A
Non-ECE Member: Haonan Wang
Member 3: Indrakshi Ray
Addional Members: Ryan Kim
•B. W. K. Hung, S. R. Muramudalige, A. P. Jayasumana, J. Klausen, R. Libretti, E. Moloney, and P. Renugopalakrishnan, “Recognizing Radicalization Indicators in Text Documents Using Human-in-the-Loop Information Extraction and Natural Language Processing Techniques,” To be presented at the IEEE International Symposium on Technologies for Homeland Security, Woburn, MA USA, Nov. 2019.

•S. R. Muramudalige, B.W. K. Hung, A.P. Jayasumana and Indrakshi Ray, “Investigative Graph Search using Graph Databases,” Graph Computing (GC 2019), Laguna Hills, CA, Sept. 2019.

•S.R. Muramudalige, H.M.N.D. Bandara, and N. Samarasekara, “Simulated Annealing Based Optimized Driver Scheduling for Vehicle Delivery,” M.Sc. thesis, University of Moratuwa, Sri Lanka, 2018.

•S.R. Muramudalige, H.M.N.D. Bandara, “Automated Driver Scheduling for Vehicle Delivery”. In: Kováčiková T., Buzna Ľ., Pourhashem G., Lugano G., Cornet Y., Lugano N. (eds) Intelligent Transport Systems – From Research and Development to the Market Uptake. INTSYS 2017. Lecture Notes of the Institute for Computer Sciences, Social Informatics and Telecommunications Engineering, vol 222. Springer, Cham, 2018. doi : 10.1007/978-3-319-93710-6_23

•S. Muramudalige and H.M.N.D. Bandara, “Demo: Cloud-Based Vehicular Data Analytics Platform”, in Proc. 14th Annual International Conference on Mobile Systems, Applications, and Services (MobiSys 2016), Singapore, June 2016. doi : 10.1145/2938559.2948849

•M. Amarasinghe, S. Muramudalige, S. Kottegoda, A.L. Arachchi, H.M.N.D. Bandara, and A. Azeez, “Cloud-based Driver Monitoring and Vehicle Diagnostic with OBD2 Telematics”, International Journal of Handheld Computing Research (IJHCR), vol. 6(4), pp. 59-75, 2015. doi : 10.4018/IJHCR.2015100104

•M. Amarasinghe, S. Kottegoda, A.L. Arachchi, S. Muramudalige, H.M.N.D. Bandara, and A. Azeez, “Cloud-based Driver Monitoring and Vehicle Diagnostic with OBD2 Telematics”, in Proc. IEEE 15th International Conference on Advances in ICT for Emerging Regions (ICTer), 2015. doi : 10.1109/ICTER.2015.7377695
Program of Study: