An Internet of Things Botnet Detection Model Using Regression Analysis and Linear Discrimination Analysis

The Internet of Things (IoT) has become a hot area of research in recent years due to the significant advancements in the semiconductor industry, wireless communication technologies, and the realization of its ability in numerous applications such as smart homes, health care, control systems, and military. Furthermore, IoT devices inefficient security has led to an increase cybersecurity risks such as IoT botnets, which have become a serious threat. To counter this threat there is a need to develop a model for detecting IoT botnets. This paper's contribution is to formulate the IoT botnet detection problem and introduce multiple linear regression (MLR) for modelling IoT botnet features with discriminating capability and alleviating the challenges of IoT detection. In addition, a linear discrimination analysis (LDA) model for distinguishing between normal activities and IoT botnets was developed. Network-based detection of IoT (N-BaIoT) dataset was used to evaluate the performance of the proposed IoT botnet detection model in terms of accuracy, precision, and detection rate.  Experimental results revealed that the proposed IoT botnet detection model provides a relevant feature subset and preserves high accuracy when compared with state-of-the-art and baseline methods, particularly LDA.