Hybrid Deep Learning Technique for Cybersecurity Detection and Classification

Abstract

Nowadays, cyber threats (CT) evolve rapidly, and this necessitates developing strong and intelligent prediction models that are effective for the detection and classification of cyber security (CS). Hence, a new Elman Crayfish network (ECFN) is proposed to predict and classify CT. In this study, a Kaggle CS threat dataset is trained with Python to develop a more effective classification model. The dataset undergoes a data refinement stage, where noisy data is preprocessed to improve precision. In order to effectively choose the features, a Crayfish Optimization Algorithm is applied in a spatiotemporal feature analysis to select the relevant attributes that contribute to classification. The ECFN utilizes these chosen features to predict CT more effectively. Finally, the detected attacks are classified, and the performance is measured to obtain high accuracy and reliability in detecting CT. The developed method improves CS protection by optimizing the selection process and improving the accuracy of classification. The model's performance is evaluated with metrics like F score, accuracy, recall, precision, and error rate, and the comparison of the results with existing approaches proves its efficiency.