This paper presents a method for accurate traffic classification using a Naïve Bayes estimator, leveraging hand-classified network data. The authors apply this technique to categorize network traffic by application, achieving high accuracy. The study demonstrates that the simplest Naïve Bayes estimator can achieve about 65% accuracy in per-flow classification, which improves to over 95% with refined variants. The refined methods include kernel density estimation and feature selection using Fast Correlation-Based Filter (FCBF). The results show significant improvements over traditional techniques, which typically achieve 50-70% accuracy. The paper also highlights the importance of using well-known traffic data for training and testing, allowing the classification of traffic using commonly available information alone. The authors emphasize the practicality and effectiveness of their approach, which can be applied to various network activities, including security monitoring, accounting, and Quality of Service.This paper presents a method for accurate traffic classification using a Naïve Bayes estimator, leveraging hand-classified network data. The authors apply this technique to categorize network traffic by application, achieving high accuracy. The study demonstrates that the simplest Naïve Bayes estimator can achieve about 65% accuracy in per-flow classification, which improves to over 95% with refined variants. The refined methods include kernel density estimation and feature selection using Fast Correlation-Based Filter (FCBF). The results show significant improvements over traditional techniques, which typically achieve 50-70% accuracy. The paper also highlights the importance of using well-known traffic data for training and testing, allowing the classification of traffic using commonly available information alone. The authors emphasize the practicality and effectiveness of their approach, which can be applied to various network activities, including security monitoring, accounting, and Quality of Service.