A Novel Method for Detecting Telecom Fraud User

Telecom fraud has been a worldwide problem with substantial annual revenue losses for many subscribers and service provider companies. To achieve effective and inexpensive detection for telecom fraud users, we propose an effective and applicable fraud user detection method based on user's Call Detail Record (CDR). The proposed method consists of two modules, namely machine learning module and template detection module. In the machine learning module, a Support Vector Machine (SVM) algorithm based on supervised learning is used to classify users using summary characteristics. In the template detection module, a Finite State Machine (FSM) based on fraud user's behavior is used to filter suspicious users, which generated in the machine learning module. After the two modules, fraud users can be detected. We implement our approach and evaluate it on a real-world dataset. The experiments show that the method can achieve high detection accuracy of 93.56%, which demonstrate that the proposed method has more excellent performance in comparison with the state-of-the-art approaches.