MultiMAuS is an agent-based simulator for payment transactions, intended for the analysis and development of dynamic on-line fraud detection methods via a multi-modal user authentication system. The multi-modal authentication procedure allows for a flexible number of authentication steps a user has to do before a transaction is processed (or rejected). It can thus adapt to the risk associated with a certain transaction, in the context of a given user. Our simulator is based on real-world credit card transaction data, to realistically model customer behaviour. The simulator can be used to study short and long term consequences of fraud detection algorithms, for different scenarios like varying levels of fraud or authentication steps. The implementation was done in Python, and is publicly available together with aggregated real transaction data (which serves as input to the simulator) and an example simulated transaction log. Implementation available: http://github.com/lmzintgraf/MultiMAuS
The 29th European Modeling and Simulation Symposium-EMSS, Barcelona, Spain, 2017.