In modern payment systems, the user is often the weakest link in the security chain. To identify the key vulnerabilities associated with the user behavior and to implement a number of measures useful to protect the payment systems against these kinds of vulnerability is a real hard task. To this aim, we designed an architecture useful to divide the users of a payment system into pre-defined classes according to the type of vulnerability enabled. In this way, it is possible to address actions (information campaigns, alerts, etc.) towards targeted users of a specific group. Unfortunately, the data useful to classify the user typically presents many missing features. To overcome this issue, a tool was developed, based on artificial intelligence and adopting a meta-ensemble model, to operate efficiently with missing data. Each ensemble evolves a function for combining the classifiers, which does not need of any extra phase of training on the original data. The approach is validated on a well-known real dataset of Unix users demonstrating its goodness. Copyright © 2017 for this paper by its authors.
Add the full text or supplementary notes for the publication here using Markdown formatting.