Smart Cyber Intrusion Detection System Using Hyperparameter Tuned Deep Learning with Feature Subset Selection Model

• Published in: 2023 6th International Conference on Engineering Technology and its Applications (IICETA) pp. 807 - 813
• Main Authors: Hadi, Maysam Reyad, Hassan, Ahmed R., Mohammed, Ibraheem Hatem, Alzubaidi, Laith H., Obaid, Mohammed Kadhim, Mohammed, Aymen
• Format: Conference Proceeding
• Language: English
• Published: IEEE 15-07-2023
• Subjects: Benchmark testing; Computer hacking; Cyberattacks; Cybersecurity; Deep learning; Feature extraction; Intrusion detection; Metaheuristics; Noise reduction; Transforms
• ISSN: 2831-753X
• DOI: 10.1109/IICETA57613.2023.10351424

Since drastic increase in network connectivity and a considerable number of computer oriented applications in recent times, the challenging issue of satisfying cybersecurity gets raised. It is needed to design an effective defence system for several cyberattacks. Therefore, the detection of instability and attacks requires the development of intrusion detection systems (IDS) to accomplish cybersecurity. Artificial intelligence (AI) techniques, specifically, machine learning (ML) and deep learning (DL) can be employed for the design of effectual data-driven IDS models. This manuscript introduces a Smart Cyber Intrusion Detection System using Hyperparameter Tuned Deep Learning Model with Feature Subset Selection (HTDLM-FSS). The presented HTDLM-FSS technique focuses majorly on the detection and classification of cyber intrusions in the smart environment. To transform the input data into meaningful format, the initial stage of data pre-processing is carried out. Then, the presented HTDLM-FSS technique accomplishes feature selection process using the wolf pack optimization (WPO) algorithm. Next, modified denoising autoencoder (MDAE) is employed for cyber intrusion recognition process. At last, gravitational search optimization (GSO) algorithm is utilized for effectual hyperparameter adjustment process. The stimulation analysis of the HTDLM-FSS technique can be tested by using a benchmark dataset and the outcomes implied the promising performance of the HTDLM-FSS approach.