DEIGASe: Deep Extraction and Information Gain for an Optimal Anomaly Detection in IoT-based Smart Cities

Abstract

Abstract A smart city architecture involves the integration of information and communication technology with gadgets across a system in order to boost connectivity for residents. As a result of ongoing data collection to improve service to customers. With the availability of multiple devices and remote flow through channels, the probability of cyber-attacks and intrusion detection has increased. As a consequence, numerous solutions for securing IoT have been implemented, including authentication, availability, encryption, and data integrity. Intrusion detection systems (IDSs) are an effective cyber solution that could be expanded by utilizing machine learning (ML) and deep learning (DP) techniques. This study presents an enhanced IDS that makes use of This study provides an optimal anomaly detection model, called DEIGASe which combines deep extraction based on the stacked autoencoder and feature selection utilizing Information gain (IG) and Genetic algorithms (GA) for select best features. The proposed model was evaluated on the upgraded IoT-23, BoT-IoT, and Edge-IIoT datasets using the GPU. When compared to existing IDS, our approach provides good ACC, recall, and precision rating performance features, with over 99.9% on record detection and calculation times around 17s for learning and 0.613s for detection.