Abstract
The deployment of heterogeneous IoT devices and the traffic generated from them creates a larger landscape in transmitting the data through the network. Often, the data in transit are sensitive and voluminous. In this context, appropriate security analysis in identifying the generated traffic alongside computationally efficient processing is required whcih is the main focus of this manuscript. So, an edge aware IoT security system is addressed using machine learning (ML), where the ML model is integrated with edge server for a reduced cloud overhead and computationally efficient processing. The use of machine learning is backed by extensive initial research demonstrating the limitations of traditional rule and signature-based security mechanisms in IoT networks. Due to the dynamic and heterogeneity of IoT traffic, static rules fail to generalize various devices and applications. Machine learning literature highlights the efficiency of data-driven models in learning baseline behavior, identifying deviations, and adapting to evolving attack patterns. Towards this, a detailed analysis between the ML model and DL model is presented.