Network Intrusion Detection for IoT-Botnet Attacks Using ML Algorithms

Amid the expanding landscape of IoT-Botnet attacks, this research delves into the augmentation of Network Intrusion Detection (NID) through the utilization of Machine Learning (ML) algorithms. Beginning with an exploration of NID's vital role in cybersecurity, this study embarks on a comprehensive investigation. Within a controlled IoT lab, in Canadian Institute for Cybersecurity (CIC), both legitimate and malicious traffic data are meticulously captured and subjected to thorough analysis. Simulated attacks, executed via Kali Linux tools, generate datasets converted into csv formats for systematic examination. The devised methodology encompasses an array of stages, including meticulous data pre-processing, effective partitioning, prudent feature scaling, and the rigorous assessment of diverse ML algorithms-namely, Decision Tree, Random Forest, Logistic Regression, KNeighbors Classifier, and Gaussian Naïve Bayes. Through meticulous training and meticulous testing, the Decision Tree Classifier emerges as a standout performer, demonstrating an impressive accuracy rate of 99.17%. Close behind, the Random Forest Classifier exhibits a commendable accuracy of 99.11%, while the KNeighbors Classifier achieves a noteworthy 98.22% accuracy. These outcomes underline the profound potential of ML algorithms in not only identifying but also effectively countering the escalating challenges posed by IoT-Botnet attacks. The cumulative results of this study contribute substantially to the reinforcement of IoT network security, effectively safeguarding against the ever-evolving landscape of threats.