Enhancing Network Security Through Advanced Intrusion Detection: A Fusion of Dimensionality Reduction and Machine Learning Classification for Improved Accuracy
J Suriya Prakash, Priya Gandhi, Dilip Kumar, Mukul Vishwas, Yash Shah · 2024
One of the biggest issues facing advancements in network security is accurately identifying and stopping breaches. Using the Friday-Working Hours-Afternoon-DDos.pcap_ISCX11 dataset, this study aims to determine which algorithm is more accurate in detecting intrusions, more especially Distributed Denial of Service (DDoS) attacks. The dataset was selected because it is relevant to real-world circumstances, particularly those involving Friday afternoon work hours. A comparison analysis is carried out to find out how well different classification algorithms perform in terms of intrusion detection. The project entails tedious performance evaluation, substantial parameter tweaking, and algorithm experimentation. Finding the algorithm that best handles the dataset’s distinctive patterns is the goal; eventually, this will help intrusion detection systems evolve. The goal of this study’s findings is to offer helpful details about the best intrusion detection system for the given dataset. The moto of this research is to identify the insights and performance in classification using different algorithms along with the DR algorithm for IDS. Here we used 6 different machine learning algorithms to identify the best accuracy using the dataset Friday-Working Hours-AfternoonDDos.pcap_ISCX11, where Ada Boosting Classifier algorithm gave the best accuracy of 99.984.The code and the Algorithms are available at: https://github.com/TheDeadpool007/IDS-using- ML