Identity resolution for fraud prevention

Abstract

The purpose of this master thesis is to find the best solution for identity resolution between social media networks and bank customers. With the contacts of the identified social media user it should be possible to identify fraud based on circular references of money transactions.The literature shows that identity resolution between multiple offline data sources is well researched. With the emerging social media networks the identity resolution between these networks was also deeply researched. Both problems can be solved by existing solutions. However, identity resolution between offline data and social media networks is not well researched, yet. The thesis is based on design research. In the different phases knowledge was gained trough literature research, interviews and meetings. An artefact was developed to evaluate and optimise the different algorithm variations. The final algorithm was then evaluated with a set of test data. In this work an algorithm is presented, which is able to identify social media users based on bank customer information with an accuracy of 80%. The key to a successful identity resolution lies within the similar data structure of the money transactions of a bank customer and the contacts of a social media profile. The best identity resolution was achieved with different weighting for the different attributes and by the normalisation of the transactions in addition to the normalisation of the name based on the name frequency.The conclusion of this work is that the thesis statement is confirmed. It is possible to correctly identify a person within a social media network based on the information availablefrom a bank customer.