A self-adaptive system for automated multi-protocol identification based on LDA
Luyan Chen, Yuanan Liu · 2016
Identifying network traces associated with multi-protocols is an essential tool for network and system security. The state-of-the-art traffic classification methods based on payload of network packets can achieve high accuracy. However, update model parameters according to current network environment to achieve the best classification performance is time-consuming and laborious. In this paper, we propose SMISL, a self-adaptive multi-protocol identification system based on LDA, which adopt the adaptive optimal model selection algorithm based on improved Genetic Algorithm (GA) to choose optimal system parameters. In order to validate the effectiveness and accuracy of SMISL, several experiments over four application protocols (SMTP, DNS, POP3 and HTTP) have be conducted, the results showed that average precision and recall is up to 99%.