Enhancing network security through machine learning: a comparative study of classification algorithms
Khushboo Tripathi, Saurabh Das · 2025
Accurate network traffic identification and categorization are crucial in cyber security to guard against various risky activities. In order to categorize the network traffic into different types of cyber-attacks, this study proposes a machine learning technique using the Weka software. For this study, NF-UNSW-NB15–v2_Preprocess data, that contains 1,048,575 instances and 21 features was used. For classification, random forest technique was utilized and 10-fold cross validation has been used to achieve an overall accuracy of 99.886%. This study shows the efficacy of machine learning algorithms for distinguishing between illicit and safe network activities. Using the reference dataset for network intrusion detection system (NIDS), the resilience, speed and accuracy of random forest is evaluated in identifying network breaches. Preliminary findings suggest that network security can be enhanced by using the random forest approach. However, the efficacy of the method may differ based on the kind of network traffic and the type of intrusion. The proposed study concludes by discussing the usefulness of machine learning techniques, in particular the random forest, in identifying the network breaches accompanied by the suggestions for additional research.