Intelligent agent based virtual clustering and multi-context aware routing for congestion mitigation in secure RPL-IoT environment

Abstract

RPL (Routing Protocol for Low power Lossy network) is a routing protocol for Internet of Things (IoT) network with low power consumption. RPL faces many security challenges because of its resource constrained nature. That security threats introduced heavy congestion in the network. If the congestions are not properly managed then it will reduce overall network performance, in addition it also reduces the network lifetime. In real time IoT application, the parent nodes attacks massive amount of nodes and establish the unbalanced structure that reduces network reliability. In this research we proposed Secure RPL (Sec-RPL) to perform congestion control RPL-IoT environment. For that we proposed five phases, (1). RPL node authentication initially, all the nodes are register their ID, IP, MAC, Rank and PUF in the sink node which generate secret key for that node by blowfish algorithm, and hash values are generated by lightweight hash one algorithm. (2). To increase network reliability we propose multi-context aware parent selection, the context decisions are made by grey set theory based decision making algorithm. For optimal parent selection we deploy Hybrid Cuttle Fish-Harris Hawker Optimizer (HC-HHO) algorithm. (3) Congestion detection and mitigation, for congestion monitoring we propose Multi-Intelligent Agent based Q-Learning (MIAQ) which mainly relies on three environmental status such as buffer status, generation status and loss status. For congestion mitigation we proposed virtual cluster based congestion mitigation (VC2M) method. (4). Deep packet analysis and attack detection, Lightweight Convolutional Neural Network with Transfer Learning (LiteCNN-TL) algorithm is deployed to analyze each packet in depth to detect malicious packets. (5). Adaptive Trickle Timer, Dynamic Rule base Trickle (DR-trickle) is deployed to adaptively changed the trickle timer based on multiple parameters (Traffic Intensity, RSSI, Received Time Interval). Finally, the simulation is conducted by NS3.26 network simulator. The result shows that the proposed system achieves better performance in terms of average control traffic overhead, packet delivery ratio, average energy consumption, delay, packet loss rate ad load balancing capacity with respect to network size and number of malicious nodes.