Abstract

Although Internet of Things (IoT) devices are being rapidly embraced worldwide, there are still several security concerns. Due to their limited resources, they are susceptible to malware assaults such as Gafgyt and Mirai, which have the ability to interrupt networks and infect devices. This work looks into methods based on machine learning to identify and categorize malware in IoT network activity. A dataset comprising both malware and benign traffic is used to assess different classification techniques, such as Random Forest, XGBoost, Logistic Regression, etc., for multi-class malware detection. After a great deal of empirical testing, XGBoost comes out on top, providing 99.9% recall and accuracy. Both known and unknown malware can be detected by the trained model with remarkable precision. The use of transfer learning, where the XGBoost model is used as a basis for the rapid construction of new malware detection models, is one of the major innovations put forth. This makes it possible to quickly adjust to new dangers. More information about how real-time network traffic can be monitored with the help of the created model to find irregularities and sound the alarm. An intelligent and proactive security solution for IoT environments is offered by the machine learning technique that is being discussed. This is an efficient defense against malware because of its high accuracy, low false positive rate, real-time detection capability, and adaptability to new malware varieties changing risks associated with IoT. The suggested methods will assist in securing susceptible IoT devices and networks from obstructive malware assaults.