Understanding User Behavior in Online Banking System

摘要

Currently, online banking has become extremely popular all over the world and plays a significant role in people's daily lives. However, the user behaviors have yet to be studied carefully in existing works. In this paper, we provide a large-scale, comprehensive measurement study of online banking users based on a two-week long dataset consisting of transactions conducted by personal users in one of the top banks in China. We demonstrate the customer behaviors mostly comply with the heavy-tail distribution which implies abnormal activities. In further analysis of those activities, we figure out that most of them are generated by two types of accounts, i.e., corporate accounts paying salaries and dishonest bank employees plastering the achievement. We extract a set of features to classify the two types of abnormal accounts from the benign ones. The experimental result illustrates that our system can accurately detect them with only 0.5% false positive rate.