New security threats emerge against mobile devices as the devices' computing power and storage capabilities evolve. Preventive mechanisms like authentication, encryption alone are not sufficient to provide adequate security for a system. In this work, we propose User Group Partition Algorithm and Behavior Pattern Matching Algorithm to extract anomalous calls from mobile call detail records effectively. The system accepts the proper input of normal mobile phone call detail records as training dataset and fraud mobile phone call detail records as testing dataset. Two main processes are included in this system: grouping mobile phone calls in training dataset according to similar phone call patterns and matching the new input mobile phone call detail records with grouped mobile phone call patterns to examine the input mobile phone call detail record is normal or not. If the system detects the anomalous mobile phone behavior, the system warns the user that the suspicious mobile phone call is detected and asks the user which action will be taken.
展开▼
