Linked-Behaviors Profiling in IoT Networks Using Network Connection Graphs (NCGs)

Abstract

The internet of things (IoT) network aims to connect everything from the physical world to cyber world, and has been a significant focus of research nowadays. Precisely monitoring network traffic and efficiently detecting unwanted applications is a challenging problem in IoT networks, which forces the need for a more fundamental behavioral analysis approach. Based on this observation, this paper proposes the Network Connection Graphs (NCGs) to model the social behaviors of connected devices in IoT networks, where edges defined to represent different interactions among them. Specially, focusing on exploring connected patterns and unveiling the underlying associated relationships, we employ a set of graph mining and analysis methods to select different subgraph structures, analyze correlated relationships between edges and uncover the role feature of interaction flows within IoT networks. The experiment results have demonstrated the benefits of our proposed approach for profiling linked-behaviors and to detect distinctive attacks in IoT networks.