Security Enhancement and Attack Detection using Humboldt Squid Optimized Extended Paillier AES Encryption Scheme and Guided Attentional GAN over Internet of Things

Abstract

The Internet of Things (IoT) has seen an increase in applications due to the exponential growth of smart devices and the decline in sensor costs. Although Internet traffic tracking and categorization have been thoroughly examined over the last 10 years, this is still popular in the IoT space. This manuscript proposes a novel attack detection framework and improved encryption scheme for secure data transmission in IoT networks. Initially, the raw data samples collected from the NSL-KDD dataset are preprocessed by performing a normalization process using a variable stability scaling technique. Then, the guided attentional generative adversarial network (GAtt-GAN) technique is proposed to detect and classify the various malware attacks like U2R, DoS, R2L, Probing, normal, and unknown accurately. In addition to this, the Extended Paillier-boosted Advanced Encryption Standard (ExP-AES) technique is introduced in this framework. By taking advantage of this characteristic, the AES-encrypted keys are optimally selected using the Humboldt Squid Optimizer (HBSO) technique, preserving the transmitted data's validity and confidentiality. The method makes use of an adaptive key turnover mechanism that makes the encryption process more unpredictable and fortifies its resistance to cryptographic attacks. Various performance measures areanalyzed, like encryption time, decryption time, throughput, accuracy, F-measure, false discovery rate (FDR), reliability, security level analysis, encryption, and decryption time, and compared with the proposed framework to prove the model’s efficacy. The overall accuracy of 98.5%, F-measure of 98.07%, and FDR of 0.042, as well as encryption and decryption time of 117.5s, are obtained by the proposed framework for providing security against IoT vulnerabilities.