An Optimized Deep Learning Based Security Enhancement and Attack Detection on IoT Using IDS and KH-AES for Smart Cities

Abstract

Today, smart cities are being built with the wide deployment of the Internet of Things (IoT). Smart cities (SCs) set out in real time to ameliorate the quality of human life in respect of efficiency and comfort. Security along with privacy are the main issues in most SCs. The IoT-centric frameworks impose certain security threats on smart city applications as they are susceptible to security issues. On this account, an Intrusion Detection System (IDS) is requisite for mitigating the IoT-associated security attacks which take advantage of certain security vulnerabilities. The aim of this paper is to improve the security and attack detection rate as early as possible. In existing works, the accuracy of the attack detection rate and security are the main challenge. To overcome any drawbacks, this work proposes an IDS for detecting the IoT attacks in a city centered on the DLMNN classification. First, the sensor values from a SC are sent to the IDS system (the training phase), which is utilized for testing the respective values. Next, the preprocessing step is performed, and then feature selection (FS) is carried out with the utilization of the Entropy-HOA method. Further on, the classification using DLMNN is performed for detecting the IoT attacks. Then, the results of the classification are analyzed and the attack is identified. Next, a secure data sharing task is performed by using the KH-AES algorithm. Last, the resulting data is forecast. The weights for each layer of the DLMNN have a high impact on the classifier's output. The comparison of the existing technique and of the proposed technique with regard to FS, classification and secure data sharing reveals that the proposed technique obtained the best results.