Clustering of IoT Devices Using Device Profiling and Behavioral Analysis to Build Efficient Network Policies

Abstract

The Internet of Things (IoT) has emerged as a new paradigm, and billions of devices are connected with the internet. IoT is being penetrated in major domains of daily life like health care, agriculture, industry, smart homes and monitoring of the environment. The operator of such complex, huge and diverse heterogeneous networks may not even be fully aware of their IoT devices working, activity, behavior and resource utilization etc. The efficient management of IoT devices becomes a challenge for network managers to ensure smooth network operation. Network traffic analysis of IoT devices is a necessary and rudimentary tool to understand the behavior of devices. In this paper firstly, we identify insights of device network traffic, discuss the activity patterns of some IoT devices and present a visual description of the pattern of IoT devices. Secondly, after analyzing the device's behavior, we build and demonstrate a profile of each device based on its activity cycle and traffic patterns information. Thirdly, the K-Means clustering algorithm is used to make clusters of IoT devices using their profile information. The clustering algorithm groups similar devices in a single group. The obtained results clearly describe the patterns of devices which help the network managers to make appropriate network policies for efficient secure network management.

IoT device multi-classification using traffic behavior analysis

“…Device identification based on traffic characteristics is often based on the analysis of device traffic, so as to accurately identify device attributes. Moreover, recognition tasks are often carried out by combining multiple features [3,[8][9][10]. Traffic capture devices are set on key network nodes (such as routers and switches in the target network) to intercept data, and then the captured original traffic data is analyzed to provide a basis for identifying device attributes.…”

Context-Aware Device Classification and Clustering for Smarter and Secure Connectivity in Internet of Things

“…To classify IoT devices, various machine learning algorithms are employed, exploring diverse combinations of these feature sets [9]. Hamza et al [10] present an inventive approach to cluster Internet of Things devices using device profiling and behavioural analysis. This methodology provides a way to develop more effective network policies by categorizing devices based on their characteristics and behaviour.…”