Cybersecurity Threat Detection and Feature Selection Using Waterwheel Plant Optimization

Abstract

The increase in cybersecurity threats necessitates accurate prediction of complex security risks. Finally, this research presents a new approach by merging the Waterwheel Plant Optimization Algorithm with Long Short memory networks to enhance cybersecurity risk assessment. One hundred ninetythree countries from multiple regions have also been considered to analyze the Cyber Security Indexes dataset. An optimized binary version of WWPA performs feature selection compared to other methods. It is shown that this method successfully reduces dimensionality while maintaining high predictive accuracy. Cybersecurity trends are modeled with the help of LSTM to capture long-term dependencies in sequential data. LSTM hyperparameters are further optimized using WWPA, and the WWPA-LSTM model is developed. The experimental results show that WWPA-LSTM surpasses traditional machine learning models in how it achieves the lowest mean square error (MSE = 0.00216). The superiority of WWPA-LSTM to other optimizationbased LSTM models on predictive accuracy and model stability is validated by the comparison with others. Moreover, statistical validation using ANOVA proves that WWPA-LSTM significantly differs from other approaches and is efficient for cybersecurity risk prediction.