An IoT Device Identification Method over Encrypted Traffic Based on t-SNE Dimensionality
Jiayun Chen, Yong Zeng, Zhihong Liu, Jianfeng Ma, Tianci Zhou, Jiale Liu · 2022
Existing approaches to IoT device identification struggle to find a balance between model accuracy, computational resources and model inference speed. In this paper, we analyse the encrypted traffic of IoT devices to identify the device type and propose a t-SNE dimensionality reduction based IoT encrypted traffic device identification method. The method reduces the dimensionality of high-latitude data and retains the main features to the maximum extent. Meanwhile, various evaluation metrics are used to compare the effectiveness of the proposed algorithm with existing IoT identification algorithms. Experiments show that the proposed algorithm achieves more than 98% accuracy on IoT device data using decision tree and random forest models, while largely improving the efficiency of model inference, outperforming comparison algorithms, and meeting realtime requirements.