Use of social network in detecting financial frauds

1. USE OF SOCIAL NETWORK
IN DETECTING
FINANCIAL FRAUDS
SARTHAK DASGUPTA
RAHUL KUMAR SHIRNET
ROHIT KUMAR SINGH

2. INTRODUCTION
➢ Financial fraud happens when someone misleads you into some malicious and deceptive scheme
and harms your financial well being by acquiring important details to take your money. Most
victims who loose their money are unable to recover it. That is why taking serious measures to
curb this kind of activities has become imperative.
➢ The impact of fraud on organizations is becoming increasingly costly. Every year financial
institutions lose millions of dollars in revenue to systematic fraud. The emergence of new
technologies and forms of payments, as well as sophistication in fraud, complicate the challenges
faced by organizations in creating effective fraud detection strategies.

3. CLASSICAL APPROACH
The ability to link multiple data sources, analyze large volumes of data, and apply newer algorithms on
the transactions, provide organizations an opportunity to capture, and sometimes predict, fraud in a
more efficient manner. This is called the classical approach. The classical approach to fraud
identification relies on creation of explicit rules (IF-THEN-ELSE IF-…) based on the recommendation
of experts. These rules are developed and modified through their collective field experiences..

4. SOCIAL NETWORK ANALYSIS METHOD
➢ The optimal way to stop such crimes is to discern them before it finds its next victim. To do so
FRAUD DETECTION and ANALYSIS comes into picture. It is imperative for the investigator to
draw out the network using such connections to get to the root of this fraud. Rarely does an
investigator look across product lines to identify fraudulent connections.
➢ However, with the introduction of social network analysis (SNA), investigators are now able to
detect data patterns. Social Network Analysis is a technique which represents the entities as nodes
and relationships between the entities as links. Representing the relationships reveals a lot more
information than simply listing out the properties of the entities.

5. SOCIAL NETWORK ANALYSIS METHOD
At a basic level, a social network consists of nodes (vertices) that are connected to other related
nodes by links (relationships). The connection between two nodes is called an edge. If all the nodes
in a social network are connected to each other, it is called a fully-connected network. A path refers
to a collection of nodes that are connected by a link. The diagrams below show three simple
variations of a financial fraud social network

6. SOCIAL NETWORK ANALYSIS METHOD
Density and Centrality
● Density is the general level of linkage among the social network nodes. It is defined as the
number of edges in a portion of a social network to the maximum number of edges that
theoretically make up the social network.
Density= C/(n (n-1)/2)
● The following formula is used to calculate the relative density of a portion of a social network.
C is the number of observed edges and n the total number of nodes in the social network. n(n-
1)/2 gives us the theoretical maximum number of edges that are possible in a given social
network.
Density measures can take any value between [0,1], with 0 representing the least
density and 1 the maximum density.

7. SOCIAL NETWORK ANALYSIS METHOD
Centrality
Centrality is the measure of closeness of a node to the center(s) of high
activity in a network and implies the structural importance of the node in the
network. Centrality can be measured by degree, closeness and betweenness
❖ Degree is the direct count of the number of connections a node has to
other nodes. A higher degree value compared to peers denotes higher
influence in the network.
❖ Closeness focuses on the overall closeness of a node to all other nodes
in the network. The closeness of a node gives it quicker access to all
other nodes in the network.
❖ Betweenness measures the extent of a node’s placement on the shortest
path between other pairs of nodes in the network.

8. HOW SNA IS IMPLEMENTED
A Bank specializes in financing vehicle dealers and charges an interest whose
rate progresses from 3, 6, 12, 24 and 36 months for each financed account
Bank Observations:
● The bank identified that most dealers closed their accounts between 3rd and
the 6th month.
● Further investigation showed that the cars associated with the closed
accounts appeared again in new accounts opened by other dealers following
a short time gap.
● It became obvious that there was an elaborate system of collaboration
between the dealers to avoid paying higher interest rates

9. HOW SNA IS IMPLEMENTED
1. Nodes: These are the suspect dealers which are selected based on the criteria
mentioned.
1. Linkages: This was recursively expanded using the snowball method to include all their
accounts and the cars involved.
1. Clusters: After diffusing the account and car nodes, the fraudulent relationships between
dealers were identified as clusters.

10. HOW SNA IS IMPLEMENTED
How SNA was used here
1. Density measures were applied to the network component to identify the clusters with
high fraudulent activity.
1. The identified clusters were subjected to centrality measures to identify key actors
within each cluster.
1. As a mitigating measure the bank took a proactive step to confront the involved
dealers and break up the identified fraud clusters.
1. A new dealer can now be easily placed in a cluster depending on his fraudulent
nature

11. NEO4J GRAPH ALGORITHM
NEO4J Graph algorithms provide one of the most potent approaches to analyzing
connected data because their mathematical calculations are specifically built to operate
on relationships. Its Data science library consists implementation of:
● Path Finding - these algorithms help find the shortest path or evaluate the availability and
quality of routes
● Centrality - these algorithms determine the importance of distinct nodes in a network
● Community Detection - these algorithms evaluate how a group is clustered or
partitioned, as well as its tendency to strengthen or break apart
● Similarity - these algorithms help calculate the similarity of nodes
● Link Prediction - these algorithms determine the closeness of pairs of nodes
● Node Embeddings - these algorithms compute vector representations of nodes in a
graph.
● Node Classification - this algorithm uses machine learning to predict the classification of
nodes.

12. THANK YOU