Group-Group Interactions D.B. Skillicorn 1 , F. Spezzano 2 , V.S. - - PowerPoint PPT Presentation

Understanding South Asian Violent Extremist Group-Group Interactions

D.B. Skillicorn1, F. Spezzano2, V.S. Subrahmanian2, M. Garber2

1Queen’s University, Kingston, Canada 2University of Maryland, College Park, MD, USA

Contact Info: vs@cs.umd.edu

1

The 2014 International Symposium on Foundations of Open Source Intelligence and Security Informatics Beijing, China, August 18-19, 2014

Motivations

• Terrorist groups cooperate each other, with an

increasing risk of terroristic acts.

• Understanding inter-group interactions is

critical for security and intelligence agencies.

• We studied the case of terrorist groups

interactions in the area of South Asia.

2

Motivations

Which of these groups play a central role in providing financial, logistical, operational, and political support to other SAVE groups?

3

SAVE Analysis

• We built a dataset (graph) describing a set of 500

interactions (edges) between 30 SAVE groups (nodes) over a period of 20 years (1993-2013).

• We used the following techniques to analyze the

SAVE network:

• 1. Spectral Embedding of Graphs
• 2. Simmelian Ties and Backbones
• 3. PageRank
• 4. Betweenness Centrality

4

The SAVE Dataset

• Our dataset describes the interactions of 30 South Asian (India, Pakistan,

Afghanistan, Bangladesh) Violent Extremist groups from 1993 to 2013. – Some of the groups included are: Al Qaeda (AQ), Lashkar-e-Taiba (LeT), Taliban (Tal), Tehreek-e-Taliban Pakistan (TTP), etc. – Our dataset also includes

• the Pakistan’s Inter-Service Intelligence agency (ISI),
• the Pakistani Islamist political group Jamaat-e-Islami (JeI), and
• some political parties.

(non-terrorist groups providing political, military or economic support)

5

The SAVE Dataset: Example Tuples

(LeT, Logistical, Al Qaeda, 1999, 2001) (ISI, Financial, LeT, 1990, 2013) (SIMI, Operational, LeT, 2006, 2006) (ISI, Political, D-Company, 1994, 2010)

• We documented a total of 500 interactions for 20 years (1993-2013)
• All the SAVE dataset was collected manually and from open sources:

– books, government publications – scholarly papers, NGO reports, dissertations, – New York Times and Times of India – Long War Journal and South Asia Terrorism Portal

6

Spectral Embedding of Graphs

• Embedding the undirected version of the SAVE

network in a 2 or 3 dimensional space to

– cluster groups that are similar, and – identify groups that are central.

7

Spectral Embedding of Graphs

• We used the Symmetric Laplacian matrix of the SAVE network A

L = D−1/2 A D−1/2

where – A(i,j)=1 if there is an edge from i to j, A(i,j)=0 otherwise (adjacency matrix), and – D(i,j)=Σj in [1:n] A(i,j) is the diagonal degree matrix.

8

Simmelian Ties and Backbones

• We constructed Simmelian networks from the undirected version of

the SAVE network.

• Triads in the social network: continuity if a node fails, mediation in

case of disputes.

• We also combined the Simmelian network with the spectral

embedding.

9

PageRank Centrality

• The PageRank PR of a node v is a measure of

the influence exerted by v

• For computing PageRank, we reverted the

directed edges so that edges go from a recipient of financial, logistical, operational, or political support to the provider.

– In this way, provider nodes have high PR score

10

Betweenness Centrality

• The betweenness centrality BC of a node v is the

percentage of shortest paths between any pair of nodes (s,t) that pass through the node v

• Nodes with high BC represent distributors of the

type of support involved (financial, logistical, political, or operational).

11

ANALYSIS OF THE WHOLE SAVE NETWORK

12

Analysis of the whole SAVE Network

3D Spectral Embedding of the undirected SAVE graph

• Closeness corresponds to

similarity (in interacting with all other groups

• approx. in the same way)
• The most important

groups are placed closed to the center

13

Analysis of the whole SAVE Network

3D Spectral Embedding of the undirected SAVE graph

• Closeness corresponds to

similarity (in interacting with all other groups

• approx. in the same way)
• The most important

groups are placed closed to the center

14

There are 3 visible clusters in the figure

Analysis of the whole SAVE Network

3D Spectral Embedding of the undirected SAVE graph Most Important Groups:

– Al Qaeda (AQ), Taliban (Tal), Jaish-e-Mohammed (JeM), Harkat-ul-Jihad-al-Islami (HuJI), Lashkar-e-Taiba (LeT), Pakistan’s Inter-Service Intelligence (ISI), Indian Mujahideen (IM), and Rabita Trust (RTr)

15

Rabita Trust serves as an important source of funds to many SAVE groups

Analysis of the whole SAVE Network

3D Spectral Embedding of the Simmelian graph S9

HuJI JeM Tal AQ LeT ISI SIMI 16

The SAVE central core is a clique of 5 groups: ISI, JeM, LeT, Tal and AQ The analysis confirms the widely held belief that the ISI cooperates with many of these groups

FINANCIAL SUPPORT ANALYSIS

Public evidence that one group provided monetary support to another

• rganization for a particular attack or for the organization’s general work

17

Financial Support Analysis

• PageRank: measures the

importance of a (financial) provider node

18

SPIDER GRAPH

The node size captures its PageRank importance The higher a node’s centrality, the closer it is to the center Edges are the group (financial) interactions

Financial Support Analysis

• PageRank: measures the

importance of a financial provider node

19

The key nodes involved in providing financial support are:

Al Akhtar Trust (AlAT) Lashkar-e-Islami (LeI) Tehreek-e-Taliban Pakistan (TTP) Haqqani (Haq) Al Qaeda (AQ)

Financial Support Analysis

• Betweenness Centrality:

how funds are distributed

20

All the above first four groups (Haq, AQ, TTP, LeI) are significant players in also in distributing financial support. AlAT is important in providing funds, while Dco in distributing them.

LOGISTICAL SUPPORT ANALYSIS

Support for the organization’s general operations from one organization to another

21

Logistical Support Analysis

• PageRank: measures the

importance of a logistical provider node

22

The dominant logistical support providers are:

Lashkar-e-Taiba (LeT) Pakistani Intelligence Agency (ISI) Other importants players: Taliban (Tal) Student Islamic Movement of India (SIMI) Al Qaeda (AQ)

Logistical Support Analysis

• Betweenness Centrality:

how logistical support is

distributed

23

The Betweenness Centrality spider graph shows that AQ, LeT, Tal, and ISI are very important in distributing logistical support, too.

OPERATIONAL SUPPORT ANALYSIS

Operational support in carrying out a particular attack from one

• rganization to another

24

Operational Support Analysis

• Spectral Embedding of the SAVE sub-

network containing operational interactions only.

SIMI HuJIB HuMuj DCo ISI IM LeT JeM Haq HuJI AQ Tal SSP LeJ AlAT AlRT TTP LeI

25

From the graph it is clear a strong cooperation between LeT, IM, AQ, Tal, LeJ (Lashkar-e-Jhangvi), and HuJI (Harkarat-ul-Jihad-al Islami)

Operational Support Analysis

• PageRank measures the

importance of a operational provider node

26

The dominant operational support providers are: Lashkar-e-Taiba (LeT) Harkat-ul-Jihad-al Islami (HuJI) Taliban (Tal) Indian Mujahideen (IM) Pakistani Intelligence Agency (ISI) Al Qaeda (AQ) It is well known that HuJI participated in many attacks with other organizations

Operational Support Analysis

• Betweenness Centrality

how operational support is distributed

27

The key players are: Lashkar-e-Taiba (LeT) Taliban (Tal) Tehreek-e-Taliban Pakistan (TTP) Al-Akhtar Trust (AlAT) Harkat-ul-Jihad-al Islami (HuJI) Al-Akhtar Trust (AlAT), typically a financing organization, is also an active

• perational player

POLITICAL SUPPORT ANALYSIS

When an organization or its leader publicly praised an other group, or the existence of ties or meetings is known

28

Political Support Analysis

• Spectral Embedding of the SAVE sub-

network containing political interactions only.

LeO AlAT AlRT PML UTeN JeM SSP LeI PPP LeJ TTP HuMuj Tal Haq JUeI AQ HuJI ISI DCo APHC JKLF LeT HuJIB HizMuj SIMI IM ISYF JeI RTr

29

The Pakistani Intelligence Agency (ISI) is the center of the graph, while other groups are strongly divided into 3 distinct clusters

Political Support Analysis

• PageRank measures the

importance of a political provider node

30

Al Qaeda is the most important provider

• f political support.

Other important players are: Jamaat-e-Islami (JeI), SIMI, TTP, ISI, LeT, and Lashkar-e-Jhangvi (LeJ)

Political Support Analysis

• Betweenness Centrality

how political support is distributed

31

The BC spider graph shows that Al Qaeda, ISI and TTP play also a significant role in distributing political support. LeJ and Jamaat-e-Islami (JeI) have also an important role in political support

CONCLUSIONS

32

Conclusions

• Our analysis is summarized in the following

tables:

33

Number of times the group appears in PR or BC top-list (rank, value)

Conclusions

– Al Qaeda is the only organization active in all the 4 domains; – LeT, ISI and the Taliban are more involved in logistical, political and operational activities than in financial interactions; – TTP concentrates on financial and political support (and is strongly connected to AQ); AQ, LeT, Tal, ISI, and TTP are the main groups that represent the violent extremism in South Asia, and are supporting each other

34

35

Thanks! Any questions?

36

37