FRAUD DETECTION BASED ON CREDIT CARD TRANSACTIONS AND ON-LINE USER BEHAVIOUR

On-line payment methodologies are subjected to fraud and identity theft. Providers of payment systems have set up heuristic criteria based on experience to flag and stop suspected transactions; however, the current state-of-the-art has two main flaws:
- Fraudsters are generally able to find out new ways to bypass the newly adopted system restrictions and therefore new frauds typology may not be detected;
- With high-volume transactions is hard to develop a good fraud detection system that is at the same time cheap and that guarantees a good user experience.

Given the current state of the art, the need for an automatic, ML-based, fraud detection system is apparent; first object of this research contract will be the development of a fraud detection system through the detection of anomaly behavior of a user.

The research activity will tackle the following step:
- Identification and extraction of relevant features;
- Selection/Creation of opportune algorithms for on-line user behavior categorization;
- Cross-validation of the results;
- Study of algorithm 'real-time' functionalities. "

Sponsor: Statwolf statwolf.com
Beginning Date date: May 2016-ongoing
HIT project coordinator: Alessandro Beghi
HIT Involved Staff: Chiara Masiero
Contact: beghi[at]dei.unipd.it