Security enhancement and attack detection using optimized hybrid deep learning and improved encryption algorithm over Internet of Things

Abstract

Exponential increases in smart devices and reduced costs of sensors have increased applications using IoT (Internet of Things). There has been an extensive analysis on the Internet traffic detection and classification in the past decade, however this is still a trending subject with respect to IoT. The objective of this work is to enhance attack detection rates in a timely fashion. The security and accuracy of attack detection rates an IDS (Intrusion Detection System) that uses HCNN (Hybrid Convolutional Neural Networks) for identifying IoT attacks in a city. After completion of pre-processing stages, FS (Feature Selection) using EHOA (Entropy-Hummingbird Optimization Algorithm) is used. Subsequently, classifications that use optimizations are carried out for IoT attack detections and classification results evaluated. KH-AES (Krill Herd-Advanced Encryption Standard) algorithm in data exchanges for security. The NSL-KDD dataset is utilised in this research to implement IDS. The data was classified based on six types of attacks: U2R, DoS, R2L, Probing, normal, and unknown. The weights in HCNN layers have significant impacts on classification outcomes. The proposed scheme is compared with popular approaches in terms of FS, classifications and security of data shares where it was found that proposed approach yields commendable outcomes.

Keywords

• Security enhancement
• Attack detection
• IoT
• Entropy-hummingbird optimization algorithm (EHOA)
• Hybrid convolutional neural network (HCNN)
• Krill herd-advanced encryption standard (KH-AES)