Design of advanced intrusion detection in cybersecurity using ensemble of deep learning models with an improved beluga whale optimization algorithm

The rapid growth and evolution of the Internet over the last few decades caused more concern about frequently changing and increasing cyberattacks. The Intrusion Detection System (IDS) is a powerful tool applied in cybersecurity methods to identify and detect intrusion attacks. Also, the probability of several intrusion attacks rises with the enormous data generation. Feature selection (FS) is vital and essential for improving performance. The dataset structure can influence the efficacy of the machine learning (ML) method. Besides, data imbalance can pose a problem, but sampling techniques can assist in reducing it. As an outcome, an efficient IDS was needed to protect the data, and the innovation of artificial intelligence's (AI) sub-domains, ML, and deep learning (DL) was one of the most effective methods to deal with this issue. Therefore, this study develops an Enhanced Intrusion Detection in Cybersecurity Using Ensemble Learning with Improved Beluga Whale Optimization (IDCS-ELIBWO) technique. The proposed IDCS-ELIBWO technique mainly addresses the detection of intrusions to achieve cybersecurity in a network. In the IDCS-ELIBWO approach, the main phase of data normalization employing min-max normalization is performed. The remora optimization algorithm (ROA) is utilized for FS and diminishes computation complexity. For cybersecurity detection, the IDCS-ELIBWO technique employs ensemble learning classifiers containing three methods such as deep belief network (DBN), gated recurrent unit (GRU), and long short-term memory (LSTM). At last, an improved beluga whale optimization (IBWO) method is used for the hyperparameter tuning process. An extensive experiment is conducted to examine the improved performance of the proposed IDCS-ELIBWO method. The performance validation of the IDCS-ELIBWO technique portrayed a superior accuracy value of 99.77 % over recent methods.