A new credit card fraud detecting method based on behavior certificate

摘要

Along with the popularization of electronic commerce (e-commerce), the problem of credit card fraud become more and more serious. Therefore, the study on fraud detection is interesting and important. Extracting the cardholder's historical transaction patterns is an important way of detecting fraud. In this paper we proposes a new credit card fraud detection system (FDS) based on behavior certificate (BC) which reflects cardholders' transaction habits. In our method, the correlation between behavior features and some special cases such as festival and weekend are considered into BC. First, we extract a set of behavior features from each cardholder's transaction records. Then we construct her/his BC based on these behavior features. Finally, we compute the risk degree for each cardholder's incoming transaction based on her/his BC. If the degree is higher than a threshold, it is considered as a fraud. Our experiment shows the effectiveness of our method.