An Optimization-Based Feature Selection Approach for Accurate Attack Detection in Fog Computing

Abstract

Fog computing bridges cloud computing and IoT increasing the data processing efficiency but is dangerous for security threats. This study explores a novel approach to attack detection in fog computing environments by using the multispectral camera with the combination of the Henry Gas Solubility Optimization (HGSO) algorithm and the GrabCut segmentation technique. The proposed HGGC combines HGSO for feature selection and optimization with the Grabcut approach for the segmentation of the images to analyse multispectral data. HGSO decreases the computational complexity by selecting the optimal features, in which the Grabcut separates the doubtful patterns denote the threats, such as Distributed Denial of Service (DDoS) or unauthorized attempts. By the combination of these methods, the system establishes strong attack detection with minimum resource overhead. Experimental results demonstrate the high detection accuracy and low latency in attack identification. Consequently, the proposed approach is implemented in the PHYTHON framework. The efficiency of the developed HGGC is then demonstrated using tables and graphs by comparing the performance measures with existing methods. Accordingly, the proposed model achieved a better result by utilizing accuracy, sensitivity, specificity and false detection rate when compared to the existing methods.