Metaheuristic neural networks for anomaly recognition in industrial sensor networks with packet latency and jitter for smart infrastructures

Abstract

ABSTRACT Sensor networks are increasingly used in all aspects of the smart city infrastructures. The stream of the data generated by sensor networks contains the information which can be used to model and manage the behavior of these infrastructures. With the expansion of the scale of the sensor networks, machine learning algorithms are increasingly used for detection of anomalies in these networks. Detection of the anomalies in industrial control network should be performed in real time. An important issue in computer and industrial sensor networks is the problem of latency and jitter. In this paper, a metaheuristic algorithm for optimal detection of the unusual behaviors and anomalies in the industrial sensor networks in presence of latency and jitter is proposed. The proposed algorithm uses the recently proposed optimizers and a neural network for modeling the behavior of the sensor network. The experimental results show that the proposed algorithm is capable of recognizing the anomalies in the industrial sensor networks with high accuracy.