Fraud Graphs on AWS

Credit card fraud can occur in many ways. Fraudsters can obtain card details of people and use them for unauthorized transactions, or create synthetic identities to get multiple cards and max them out (bust-out fraud) after seemingly normal behavior. Graph applications can be used to prevent counterfeit credit card transactions by identifying common links between card holders, point of sale locations, transactions and terminals. Identifying these links in a vast network of data points can expose compromised cards which can then be blocked to prevent fraud.

Insurance companies want to identify fake claims which could slip under the radar when looking at claims in isolation. A graph application can be used to find links between claims and the people associated with each claim. This can expose fraud rings where a network of fraudsters file multiple claims, play different roles, or be involved in multiple small claims. Graphs can also be used to detect under or misrepresented data in an insurance policy based on previously known customer profile and relationships with other people and claims.

Bad actors can attempt to secure a loan for fraudulent purposes. Graph databases can explore the network of people and things connected with a person to a significant enough depth looking for any connection with other known bad actors. If such a connection is found, the loan approval applications can be flag the loan application so that banks can assess the risk and decide whether or not to approve the loan.

Fraudsters may create multiple accounts in online betting platforms, known as ‘Gnoming’. They can use these accounts to pocket the signup bonus and never play, or bet against themselves to ensure they win. A fraud graph can identify the relationships among people, e-mail accounts, IP addresses, phone numbers or postal addresses. By finding common contact information, online betting platforms can prevent gnoming.

The term “money laundering” can be defined many ways but it describes the process where illegally obtained funds are passed through various entities such that the origin of the funds is hard to trace and thus the money becomes “laundered” and appears to be legitimate income.Graph databases can help in tracking financial transactions through complex networks to provide a clear picture of the path the money took from its source to its destination.

Collusion occurs when two or more players work together to affect the outcome of a game. In online gambling, self-collusion is also possible. Self-collusion is where a person signs up for multiple accounts and plays simultaneously colluding with themselves. Graphs can be used to determine if a person has already signed up for a game by detecting common data points such as phone numbers or IP addresses.

PaySense is a digital lending startup specializing in using advanced data sciences to provide credit without the hassle of complex application processes, arduous physical proofs and long waits. PaySense unique model establishes billions of relationships between millions of people. PaySense has been an early customer since the launch of Amazon Neptune to store this model. The graph database is used to detect fraud rings that look for abnormal growth rates in certain parts of the graph.

Using Amazon Neptune, Rappi was able to use Neptune for their fraud detection solution, replace their 3rd party solution and bring down their overall cost significantly.