A Survey on Machine Learning Techniques for the Classification of Encrypted Network Traffic

Due to encrypted network traffic growth, complex machine learning (ML) algorithms are required to label applications accurately. This is because of problems connected with traditional traffic categorization in encrypted communications, especially with the use of obfuscation techniques such as VPNs. The questionnaire detected several serious issues, such as privacy problems, complexity of calculations, and applicability to unknown traffic. It finds research gaps and possible fixes, for example, combining explainable AI, adversarial learning, and uncertainty quantification to enhance classification interpretability and robustness. This survey deeply studies current machine learning (ML)-based approaches for classifying encrypted communication and evaluating their benefits, drawbacks, and suitability for practical application. This work is based on the research described in Extensible Machine Learning for Encrypted Network Traffic Application Labeling via Uncertainty Quantification.