IoT Unbalanced Traffic Classification System Based on Focal_Attention_LSTM

With the development of the Internet of Things, communication between diverse devices has become a common practice. The huge growth of IoT devices and the different characteristics of IoT traffic patterns have attracted attention to traffic classification solutions to solve the problems that arise in IoT applications. This paper proposed an IoT traffic classification system using deep learning method. The main work of our framework, Focal_Attention_LSTM, is as follows: (1) Apply the LDA feature processing algorithm. (2) Adapted the well-performed attention mechanism to the IoT traffic classification area (3) Modified focal loss is used to address common traffic category unbalances and hard example learning problems. We evaluate our system with IoT dataset BoT-IoT. The experimental results show that our system achieves 3.6% improvements compared to traditional feature processing methods and deep learning models, and has effectively solved the traffic category unbalance problem.