This document discusses whiplash for cash fraud schemes and how graph databases can help detect fraud rings. It describes how fraudsters stage fake car accidents to collect insurance payouts. Investigating claims data as a graph makes it easier to identify connections between individuals and accidents that may indicate an organized fraud ring. Queries can find people involved in multiple accidents and identify closely connected groups that deserve further investigation. Representing data as a graph allows insurers to automatically detect suspicious patterns indicative of fraud.
Driving Behavioral Change for Information Management through Data-Driven Gree...
Detecting fraud rings with graphs
1. Fraud detection and whiplash for
cash schemes
SAS founded in 2013 in Paris | http://linkurio.us | @linkurious
2. WHAT IS A GRAPH?
Father Of
Father Of
Siblings
This is a graph
3. WHAT IS A GRAPH : NODES AND RELATIONSHIPS
Father Of
Father Of
Siblings
A graph is a set of nodes linked by
relationships
This is a node
This is a
relationship
4. People, objects, movies,
restaurants, music
Antennas, servers, phones,
people
Supplier, roads, warehouses,
products
Graphs can be used to model many domains
DIFFERENT DOMAINS WHERE GRAPHS ARE IMPORTANT
Supply chains Social networks Communications
5. Stage fake accidents and receive real money
from insurance companies
WHAT IS A WHIPLASH FOR CASH SCHEME
Stage a fake car
accident
Fill insurance
claims
Cash in the
check
Based on the accident, they
fill insurance forms to ask
their insurance companies to
cover for injuries and the car
damages.
The insurance company
looks at the claim and writes
a check to its customers. The
fraudsters cash it.
A few fraudsters get together.
They define an accident
scenario and enact it.
6. But why is it hard to detect whiplash for cash
fraud rings?
WHY FRAUD DETECTION IS HARD
7. The criminal keep their claims small, corroborate
each other and pretend to have hard to disprove
injuries
PROBLEM 1 : CRIMINALS FLY BELOW THE RADAR
8. From one accident to the next, the vehicles, the
persons and their roles will change : hard to see
a pattern emerge
PROBLEM 2 : HARD TO SEE THE PATTERN IN A LARGE NUMBER OF ACCIDENTS
10. A single accident doesn’t look suspicious
A GRAPH DATA MODEL FOR A SINGLE ACCIDENT
IS_LAWYER
IS_DOCTOR
Udo
(Person)
Monroe
(Person)
Robrectch
(Person)
Skyler
(Person)
Euanthe
(Person)
Jasmine
(Person)
Chelle
(Person)
Sousanna
(Person)
Focus
(Car)
Corolla
(Car)
Accident 1
(Accident)
IS_INVOLVEDIS_INVOLVE
D
PASSENGER
DRIVER
DRIVER PASSENGER PASSENGER
PASSENGER
11. But representing the claim data as a graph
makes it easy to spot a fraud ring
WHAT DOES A FRAUD RING LOOK LIKE
3 separate accidents (above) involve a small set of 8 persons (below) who seem
to have strong relationships : suspicious?
12. HOW TO INVESTIGATE A WHIPLASH FOR CASH FRAUD RING : STARTING POINT
The investigation starts with a car accident...
As a fraud analyst, we’ll use a Neo4j graph database to investigate the claims
data and see if we can spot something suspicious
13. 1. Are the persons involved in the accident
involved in other accidents?
2. If they are, who are they involved with? Are
these people connected to other accidents?
HOW TO INVESTIGATE A WHIPLASH FOR CASH FRAUD RING : QUESTIONS
15. location date
Florida 23/05/2014
Florida 27/05/2014
QUESTION 1 : ARE THE PERSONS INVOLVED IN THE ACCIDENT INVOLVED IN OTHER ACCIDENTS
Our suspects are involved in 2 more accidents
16. With a simple “*” we are expanding our search
across the graph
QUESTION 2 : WHO ARE THEY INVOLVED WITH
MATCH (accident)<-[*]-(potentialfraudtser:Person)
WHERE accident.location = 'New Jersey'
RETURN DISTINCT potentialfraudtser.first_name as first_name, potentialfraudtser.
last_name as last_name
17. first_name last_name
Udo Halstein
Robrecht Miloslav
Monroe Maksymilian
Skyler Gavril
Euanthe Rossana
Jasmine Rhea
Sousanna Pinar
Chelle Jessie
QUESTION 2 : WHO ARE THEY INVOLVED WITH
We have a group of 8 people involved in 3
accidents
18. What if we want to detect automatically these
suspicious behaviour?
QUESTION 3 : IS IT POSSIBLE TO DETECT THE FRAUD
19. Looking in real time for highly connected
“accidentees”
QUESTION 3 : IS IT POSSIBLE TO DETECT THE FRAUD
MATCH (person1:Person)-[*..2]->(accident1:Accident)<-[*..2]-(person2:Person)-[*..2]->
(accident2:Accident)<-[*..2]-(person3:Person)-[*..2]->(accident3:Accident)
RETURN DISTINCT person1, person2, person3
20. QUESTION 3 : IS IT POSSIBLE TO DETECT THE FRAUD
It is possible to look for suspicious patterns at
large scale
An event triggers
security checks
New customer
New car registered
New accident
A Neo4j Cypher query
runs to detect patterns
Identification of the
fraudsters
21. The fraud teams acts faster
and more fraud cases can be
avoided.
WHAT IS THE IMPACT OF LINKURIOUS
If something suspicious comes up, the analysts
can use Linkurious to quickly assess the
situation
Linkurious allows the fraud
teams to go deep in the data
and build cases against fraud
rings.
Treat false
positives
Investigate
serious cases
Save money
Linkurious allows you to
control the alerts and make
sure your customers are not
treated like criminals.
26. Presentation on fraud and whiplash for cash by Philip Rathle and Gorka Sadowski (the
inspiration for this presentation) : https://vimeo.com/91743128
Article on whiplash for cash :
- the article : http://linkurio.us/whiplash-for-cash-using-graphs-for-fraud-detection/
- the dataset : https://www.dropbox.com/s/6ipfn4paaggughv/Whiplash%20for%20cash.zip
GraphGist on whiplash for cash :
- the article : http://gist.neo4j.org/?6bae1e799484267e3c60
Whitepaper on fraud detection by Philip Rathle and Gorka Sadowski :
- the whitepaper : http://www.neotechnology.com/fraud-detection/
SOME ADDITIONAL RESOURCES TO CONSIDER