Neural Network Rule Extraction for Real Time Traffic Behavior Identification

Abstract

The rapid identification of network traffic classification has become an important network management tasks. Although the current research methods can classify the network traffic and achieve high accuracy. Because of the high complexity of computing, memory consumption and other reasons, they can only provide 100 megabytes of processing power and cannot handle a large number of concurrent traffic situation. Aiming at this problem, this paper proposes a method to identify the traffic classification on the network, and selects the early identified traffic characteristics as the attributes of the training model to alleviate the resource and time consumption of the online traffic classification. The method first sends the traffic data to the trained neural network and extracts the fuzzy traffic data. And then the fuzzy traffic data optimizes by genetic algorithm and sends to the decision tree algorithm to generate decision tree. Finally, the decision tree is transformed into rules, such as if-then rules. The resulting rules use FPGA as the application context to achieve online network traffic classification. This method is not only an innovation in the field of neural network extraction, but also a novel method of neural network rule extraction to solve the online network traffic classification.