Abstract: Aiming at the difficulty of real-time detection of VoIP applications, feature extraction is performed using a time-averaged shunting rule, a classification model based on C4.5 decision tree algorithm, and a real-time packet capture detection mechanism based on Jpcap library are used to design and implement the recognition. High-precision, real-time VoIP traffic detection system. Experimental results show that under the premise of accuracy and real-time performance, the system can identify more than one kind of VoIP application program, the identification time is only 115 ms, and the accuracy is over 93%.