Automated Intrusion Detection and Classification using Binary Metaheuristics with Deep Learning on Smart Cities

With the dramatic increase of interconnected devices in smart cities, ensuring the integrity and security of the basic infrastructure has become a major concern. Intrusion detection plays a major role in protecting smart city systems against cyber-attacks and threats. Deep learning (DL) techniques, namely recurrent neural networks (RNN) or convolutional neural networks (CNNs), were employed for analyzing the network traffic data. The model was trained by labelled dataset, where network traffic instance was categorized as either intrusive or normal. This study proposes a novel automated Intrusion Detection and Classification design using Binary Metaheuristics with Deep Learning (AIDC-BMDL) techniques on Smart Cities. The AIDC-BMDL technique makes use of metaheuristic feature selection and DL based classification process. For the election of optimal features in the intrusion data, the AIDC-BMDL technique uses binary gray wolf optimizer (BGWO) algorithm. Besides, the AIDC-BMDL technique exploits Stacked Autoencoder (SAE) model for the effectual recognition and classification of the intrusions in the smart city environment. The simulation results illustrated the ability of the AIDC-BMDL technique to accurately identify intrusions in smart city environments. The AIDC-BMDL technique gained maximum performance, pointing out the significance of improving the security of smart cities.