Enhanced Security for an IoT Devices in Cyber-Physical System against Cyber Attacks

Abstract

Cyber physical systems (CPS) integrate the physical and numerical worlds. Number of computing system connected with wireless networks and also transmits sensitive data in secure manner of network. Those platforms, however, are vulnerable to cyber-attacks on people, equipment, and the ecosystem. The method of transmitting sensed data to controllers through a wireless transmission link, therefore increasing the attack surface, attracts cyber concerns in CPS. Intrusion detection systems (IDS) are well-known as powerful survival mechanisms, and machine learning (ML) approaches have lately been employed in their creation. The conventional ML-based IDS, which need a considerable computing capacity, do not work on IoT devices as computational resources such as memory, processing power and limited energy are restricted. The proposed work is to build and implement a lightweight, resource-restricted IDS-based appliance. Moreover, the presents IMPACT, a lightweight IDS model based on ML (Impersonation Attack detection). This is built on extensive object detection using a gradation sequential Support Vector Machine (SVM) to reduce the amount of variables via feature selection and extraction utilizing a C4.8 wrapper, mutual information (MI), and stacked autoencoder (SAE), to install and operate on resource-constrained devices. Using the Aegean Wi-Fi Intruder Dataset (AWID), the IMPACT taught to detect impersonator attacks. The suggested IMPACT model surpassed current state-of-the-art benchmark techniques, achieving 98.23% accuracy, 1.21% false alarm rate, and 97.65 % detection rate.