JoWUA

Abstract

A cyber-attack is the malicious manipulation of computer networks and systems to compromise data or impede procedures and operations using malware. With the exponential growth in computational capacity, machine learning (ML) and deep learning (DL) approaches have emerged as promising countermeasures for advancing and identifying such threats. To address this challenge, a novel optimized deep hybrid attack detection model called SCEEHO-SPC-CNN-CD-DBN is proposed in this research article. Data is subjected to a preprocessing procedure before it is used for further processes. Here, the data undergoes a normalizing phase for pre-processing, during which the statistics and higher-order statistical features are retrieved. The cyber-attack detection process concludes with a hybrid DL model applied to the retrieved features. The proposed hybrid classifier integrates models such as the DBN (Deep Belief Network) with contrastive divergence (CD) and the split convolution module (SPC)-based CNN (Convolutional Neural Network). Training the CNN and DBN using the SCEEHO(Sea CrowEndorsed Elephant Herding optimization) model and fine-tuning the ideal weights improves detection accuracy. Furthermore, have tested the developedSCEEHO-SPC-CNN-CD-DBN-based hybrid classifier on the CIC IoT Dataset 2023. The evaluated results, employing a wide range of statistical measures, demonstrate that the research model performs efficiently.