Gazelle-Dingo Optimization and Ensemble Classification: A Hybrid Approach for Intrusion Detection in Fog Computing

Abstract

Fog computing is a system that expands the cloud services to the network to manipulate the inherent issues of the cloud. An intrusion detection system (IDS) is a network security tool for monitoring a computer network for malicious activities that include potential threats, abnormal activities, and unauthorized access to fog networks. It is a key component of fog network security that provides quality of service in data communication. In this research, the intrusion is detected using the Ensemble classifier in the fog layer by utilizing the proposed gazelle dingo optimization (GDO). Here, three layers are considered for the entire process, including the endpoint layer, cloud layer, and fog layer. Initially, the physical process is done in the endpoint layer for connecting and exchanging the information. Then, at the cloud layer, various processes like data transformation, feature selection, and ensemble classification are carried out. The data transformation process utilizes Yeo-Johnson transformation, followed by feature selection that is done by Kulczynski similarity. Afterward, the selected features are fed toward ensemble classifiers such as deep residual network (DRN), deep belief network (DBN), and Zeiler-Fergus network (ZF-Net) for the classification, where the ensemble classifier is tuned by GDO. Finally, the fog layer detects the intrusion using an ensemble classifier tuned by GDO. The GDO is formed by combining the gazelle optimization algorithm (GOA) and dingo optimization algorithm (DOA). Here, the proposed GDO-Ensemble Classifier attained the superior values of precision, recall, and F-measure of 0.922, 0.935, and 0.918.