A Trust Aware Based Predictive Model using Hybrid Convolutional Neural Network for Mobile AD HOC Network in Internet of Things

Abstract

Mobile ad-hoc networks (MANETs) are inescapable independent Wireless Sensor Networks (WSNs) that will assume a fundamental job in upcoming trends of Internet-of-Things (IoT) communication, somewhere sharp-witted gadgets will a tendency to associated in a totally scattered manner. The IoT is a type of wireless heterogeneous network of different types such as WSNs, MANETs, Zig-Bee, WI-FI, and RFID. So a trust based routing in MANET based IoT network is a difficult task for better Device to Device (D-2-D) communication. Be that as it may, because of the absence of framework and the nonappearance of concentrated administration in MANETs, networks are covered with different security threats. Some inward mobile sensor nodes in these positive feature based obliged wireless networks may bargain the routing mechanism in order to attacks to do unmistakable sorts of the data packet sending mischievous activities. Methods: In order to address this type of IoT communication issue, in our previous research paper, we devised a routing protocol in IoT based on the secure and energy efficient trust aware approach using the Particle swarm based (PSO) Optimized Artificial Neural Network (ANN), it is used to classify the packet dropping adversaries before the transmission by supervising the intermediates communicating sensor nodes nature to discover route and their maintenance period. But the achieved result should be better by using the concept of Convolutional Neural Network (CNN) with PSO instead of ANN. In this paper, we perform sensitivity analysis of IoT communication using Secure and Energy Efficient Trust Aware (SEETA) routing mechanism with PSO based optimized CNN and it is used to identify the unlike parameters variation in distinctive scenarios in the existence of data packet dropping attacks or malicious nodes. Also the proposed work recapitulates the trusted route discovery mechanism with their maintenance process where routing is based on our existing SEETA protocol with the purpose of countering the certain attack or malicious patterns along with optimized CNN. Results: Simulation is conducted with MATLAB based network simulator which indicates the correct choices of parameter values for proposed IoT network scenarios. When the QoS constraints of IoT network is calculated and compared with various existing approaches, the proposed PSO based CNN with SEETA routing mechanism achieves the better performance of 99.31% in terms of data delivery rate with reduction of 16.76% in energy consumption rate as compare to exiting works. Conclusion: During simulation of proposed IoT network based on different network conditions, we observed that the achieved performance is best in terms of Energy Consumption with Throughput and Loss Rate. After that we also obtained the achieved transmission delay is less and Alive Nodes Count is more with maximum Detection Rate.