ENHANCING THE EFFECTIVENESS OF ENCRYPTED TRAFFIC CLASSIFICATION THROUGH DATA PRESERVATION AND INPUT ALIGNMENT WITH DEEP NEURAL NETWORKS

Network traffic classification plays a crucial role in network management and security. Most of the network traffic today is in encrypted form, making traffic identification more difficult. In this context, machine learning and deep learning have emerged as the foundational technologies to solve the problem. To date, numerous encrypted network traffic classifiers based on machine learning and deep learning have been proposed and extensively evaluated in experiments. However, the instability in the performance of these models when deployed on real networks has posed a challenge that has not been satisfactorily addressed so far. In this study, we propose a feasible method to build a more sustainable encrypted network traffic classifier. The classifieris builtbased on innovative input data generation techniques that preserve important latent features and facilitate the CNN deep learning network to maximise its inference ability. The proposed method aims to improve the model's performance and adapt well to the variability and resource constraints of real-world networks. Experimental results show that our model achieves classification performance comparable to state-of-the-art methods. While handling full information of the data samples to avoid missing potential variability factors, the model still maintains simplicity to minimise the limited computational cost of real networks.