Attack resilient IoT security framework using multi head attention based representation learning with improved white shark optimization algorithm

At present, the internet of things (IoT) plays a vital part in the growth of programmed electrical power stations while presenting magnificent chances, particularly cybersecurity. In IoT networks, security is now required owing to the higher amount of data to be handled. IoT cybersecurity aims to decrease the cybersecurity threat for users and organizations over protecting IoT assets and privacy. Therefore, identifying numerous anomalies or cyberattacks in a network and constructing an effectual intrusion detection system (IDS) becomes more significant. Artificial intelligence (AI), mostly machine learning (ML) and deep learning (DL), has been employed to construct a data-driven intelligent IDS. This paper presents a multi-head attention-driven intrusion detection with improved white shark optimization algorithm (MHAID-IWSOA) methodology in IoT networks. The main intention of the MHAID-IWSOA methodology relies on enhancing the cybersecurity detection and migration model using advanced optimization algorithms. Initially, the data pre-processing applies min-max scaling to transform input data into a beneficial format. Besides, the sand cat swarm optimization (SCSO) model is used for the feature selection (FS) process. The proposed MHAID-IWSOA model employs the bidirectional gated recurrent unit with multi-head attention (BiGRU-MHA) technique for attack detection and classification. Finally, the improved white shark optimization (IWSO) technique optimally alters the hyperparameter value of the BiGRU-MHA technique and results in superior classification performance. The experimental evaluation of the MHAID-IWSOA model is performed on the Edge-IIoT dataset. The extensive comparison analysis of the MHAID-IWSOA model illustrated a superior accuracy outcome of 98.28% over existing techniques.