Applying Intelligent Agents for Anomaly Detection of Network Traffic in Internet of Things Networks

Systems based on the concept of ‘Internet of Things’ (IoT) differ by multi-tiered architecture, a great number of used ‘things’, the influence of new types of attacks, the incompleteness and ambiguity of their parameters. For these reasons, solving security management tasks in IoT networks, such as network traffic analysis, requires applying intelligent approaches and methods. The purpose of the paper consists in development and assessment of a new algorithm of the network traffic analysis in a real or near real time. The paper also considers various variants for implementation of intelligent agents intended for network traffic analysis in IoT networks in different cases: (1) high-performance computers, (2) embedded devices, and (3) systems-on-chip. The agents are based on the algorithm of pseudo-gradient anomaly detection and fuzzy logical inference. The suggested algorithm operates in real time. The experimental assessment of the approach shows that the gain can reach 50% in accuracy and 90% in speed.