OPTIMIZING NETWORK SECURITY USING MACHINE LEARNING-BASED INTRUSION DETECTION SYSTEMS

This research deals with network security issues that result from the continued vulnerability of today’s computer networks to malicious attacks and unauthorized accesses. An intrusion detection model based on machine learning is developed for distinguishing between normal and malicious network data. Different types of experiments were carried out using the NSL-KDD database, which consists of training data and testing data for evaluating the results. Multiclass attacks were labeled as binary data, where normal data was labeled as 0 and malicious data as 1. Data pre-processing included data cleansing by removing unnecessary columns, one-hot encoding, and normalization of numeric attributes. Four different machine learning algorithms were compared, namely, Decision Tree, Random Forest, Naïve Bayes, and Logistic Regression. Decision Tree algorithm proved to be the most effective, with the highest accuracy of 79.62% and F1 score of 79.71%.