Improved network traffic classification using hashing techniques in machine and deep learning

The rapid global growth of the internet, driven by advancements in fiber and 5G technology, multi-device access, and affordable services, has increased the pressure on internet service providers to classify network traffic efficiently. Accurate traffic classification and protocol identification are critical for detecting malicious activity. This study introduces a new method that enhances machine learning and deep learning models by applying hashing techniques to convert string-based IP addresses into numerical values. The improved models demonstrate a significant boost in accuracy, increasing from 76% to 83%, along with better recall and F1-scores in key categories. These findings highlight the potential of hashing techniques to improve the performance of machine learning models in network traffic classification tasks.