With increasingly rampant telephone fraud activities, the social impact and economic losses caused to China have increased dramatically. Precise, convenient, and efficient fraudulent phone call recognition has become a challenge since telephone fraud became more varied and covert. To deal with this problem, many researchers have extracted some statistical features of telephone fraud behavior and proposed some machine learning algorithms on the field of fraudulent phone call recog-nition. In this paper, the calling detail records are utilized to construct a classifier for fraudulent phone call recognition. Meantime, a deep learning approach based on convolutional neural network ( CNN) is proposed for better features learning and compared with the existing state-of-the-art ma-chine learning algorithms. It learns phone number and call behavior features of telephone fraud, and improves the accuracy of classification. The evaluation results show that the proposed algorithm out-performs competitive algorithms.