Novel trust evaluation using NSGA-III based adaptive neuro-fuzzy inference system

Recently Mobile adhoc networks (MANETs) have received the great attention of researchers as these models provide a wide range of applications. But MANET nodes are prone to various security threats. To overcome this issue, many trust management frameworks have been implemented in the literature. It has been found that the use of machine learning can predict trust values more efficiently. However, machine learning performance suffers from the hyper-parameters tuning and over-fitting issues. Therefore, in this paper, novel trust management is proposed. initially, the Adaptive neuro-fuzzy inference system (ANFIS) is used to train the trust prediction model. Thereafter, a non-dominated sorting genetic algorithm-III (NSGA-III) is used to tune the hyper-parameters of the ANFIS model. Precision, recall, and root mean squared error metrics are used to design a multi-objective fitness function. Optimized link state routing (OLSR) protocol is used for comparative analyses purpose. Three different attacks are applied on the designed network i.e., link spoofing, jellyfish, and gray hole attacks to obtain the dataset. Comparative analysis reveals that the proposed trust evaluation model outperforms the competitive trust evaluation models in terms of various performance metrics such as routing overheads, average end to end latency, packet delivery ratio, and throughput. Thus, the proposed protocol is more secure against various security threats.