Enhancing Intrusion Detection using Supervised Machine Learning Algorithms

In the realm of Security of Network, an ongoing pursuit exists to counter cyber-attacks that have the potential to disrupt networks. Furthermore, the unforeseen onset and heightened reliance on the Internet have contributed to a swift escalation in malicious activities within networks. Creating a robust intrusion detection system (IDS) is crucial for identifying abnormalities in the network and thwarting unauthorized access to network resources, thereby ensuring the security of information. While several notable methods have been suggested recently to address intrusion detection challenges, constructing a secure system remains difficult due to attackers adapting their behavior to evade the system’s security mechanisms. In the present research, machine-learning classifiers such as decision tree, support vector machine, Multi-layer Perceptron, random forest, K-nearest neighbor, and logistic regression, were utilized to categorize data into normal or intrusive classes. The assessment of the model’s performance took place on NSL KDD datasets features. Acknowledging that the efficiency of IDS depends on the dimensions of the data, pre-processing was undertaken to eliminate irrelevant features from the dataset. In conclusion, the model’s overall accuracy was assessed across various attack classes, including DoS, Probe, U2L, and R2L, with the aim of identifying the most appropriate algorithm for each specific attack class.