Machine Learning Based Framework for Attack Detection on IoT Devices

The IoT devices are targeted by increased hacks due to their rapid development and widespread use. According to a report, botnet attacks dominate IoT. Due to memory and processing limitations, IoT devices face security vulnerabilities. Several rule-based detection approaches are vulnerable. Many rule-based detection systems are also attackable. This research proposes a machine learning (ML)-based botnet detection system using network traffic data from five smart home IoT devices under attack. The “Gafgyt” and “Mirai” botnet attacks against IoT devices were categorized into five. The K-Nearest Neighbor (KNN), Support Vector Machine (SVM), Naive Bayes (NB), and Random Forest (RF) ML algorithms were used to identify attack from benign traffic data in over a million records from individual devices utilizing 115 factors. The best feature analysis was done utilizing the RF method to improve model correctness and processing speed. With 99.99% accuracy, 99.87% precision, 99.79% recall, and 99.89% AUROC rating, RF is superior. Experimental results show that our approach can efficiently identify botnet attacks in network traffic.