Discovery of fraud rules for telecommunications—challenges and solutions
KDD '99 Proceedings of the fifth ACM SIGKDD international conference on Knowledge discovery and data mining
Journal of Parallel and Distributed Computing - Special issue on wireless networks
Unsupervised Profiling for Identifying Superimposed Fraud
PKDD '99 Proceedings of the Third European Conference on Principles of Data Mining and Knowledge Discovery
Detection of Mobile Phone Fraud Using Supervised Neural Networks: A First Prototype
ICANN '97 Proceedings of the 7th International Conference on Artificial Neural Networks
Dempster-Shafer Theory for Intrusion Detection in Ad Hoc Networks
IEEE Internet Computing
Detecting fraud in mobile telephony using neural networks
IEA/AIE'2005 Proceedings of the 18th international conference on Innovations in Applied Artificial Intelligence
Multi-camera people tracking using evidential filters
International Journal of Approximate Reasoning
Shape from silhouette using Dempster-Shafer theory
Pattern Recognition
Two-stage database intrusion detection by combining multiple evidence and belief update
Information Systems Frontiers
Hi-index | 0.00 |
This paper introduces a framework for fraud detection in mobile communication networks based on the current as well as past behavioral pattern of subscribers. The proposed fraud detection system (FDS) consists of four components, namely, rule-based deviation detector, Dempster-Shafer component, call history database and Bayesian learning. In the rule-based component, we determine the suspicion level of each incoming call based on the extent to which it deviates from expected call patterns. Dempster-Shafer's theory is used to combine multiple evidences from the rule-based component and an overall suspicion score is computed. A call is classified as normal, abnormal, or suspicious depending on this suspicion score. Once a call from a mobile phone is found to be suspicious, belief is further strengthened or weakened based on the similarity with fraudulent or genuine call history using Bayesian learning. Our experimental results show that the method is very promising in detecting fraudulent behavior without raising too many false alarms.