PhD Researcher in the field of data science for business decision-making
Katholieke Universiteit Leuven
Universidad Estatal, Examinar oportunidades similares
es
Universidad Estatal, Examinar oportunidades similares
This research is part of the FWO project ANUBIS: Aligned oNline and multilevelUser and entity Behavior analytics for Intelligent System security :
Digital business thrives by secure transaction processes. Despite advanced authentication procedures and network protocols, a typical organization is estimated by the Association of Certified Fraud Examiners to lose 5% of its revenues due to fraud. Developing powerful fraud detection systems that continuously monitor and learn from data flows therefore is of crucial importance to reduce losses by timely blocking, containing and preventing malicious user behavior. However, like viruses mutate in response to immunity, hackers and fraudsters continuously adapt their methods in response to organizations’ efforts to mitigate fraud. Online systems are relentlessly probed for security vulnerabilities emanating from system modifications and updates. Fraud therefore is dynamic, system-dependent and organization-specific.
Hence, a pressing need exists for adaptive fraud detection systems which on the one hand continuously adapt to system and user behavior evolutions and rapidly learn to detect new fraud patterns from the continuous stream of data that is generated by users and systems, and on the other hand align with the business needs and the organizational environment where the system is deployed. Although in the ideal world the aim of fraud detection and prevention systems is to eradicate fraud, in practice most organizations aim for pragmatic approaches that are cost-efficient in reducing fraud. This can be achieved by aligning fraud detection systems with their actual organizational role and by adapting them to the true objective, i.e., minimizing fraud losses. Therefore, in this project we will be developing systems which allow to optimize security and fraud investigation efforts in function of expected losses. To this end, systems need to learn to detect fraud involving larger losses with higher priority over fraud cases involving smaller losses, taking as well into account the cost of false alarms.
Elige el país al que quieres viajar para estudiar gratis, trabajar o hacer voluntariado
