Development of an Enhanced Intrusion Detection System Framework Using Walrus Optimization Algorithm for Support Vector Machine Parameter Tuning
The intrusion detection system (IDS) plays a vital role in modern cybersecurity by identifying unauthorized activities on a network and supporting protective responses.However, traditional IDS approaches face difficulties when dealing with complex data and sophisticated attack methods, leading to lower detection accuracy and an increase in false positives.This study introduces a new framework that combines the Walrus Optimization Algorithm (WaOA) with Support Vector Machine (SVM) classifiers to enhance the performance of SVM-based IDS, which is very sensitive to both the choice of useful feature selection and hyperparameter optimization.Suboptimal feature selection and poor hyperparameters may lead to overfitting, higher computational cost, and poor detection results.By integrating the WaOA and SVM in order to optimize SVM hyperparameters, the framework aims to boost SVM efficiency while reducing computational complexity.The CICIDS2017 dataset will be used as our input data, while the implementation will be done using MATLAB programing language.The evaluation of the experiment will be done by comparing the results of the enhanced WaOA_SVM with those of the standard SVM.The study will investigate whether hyperparameter optimization using WaOA can improve classification performance and operational efficiency compared with a standard SVM.
Authors
- Oladimeji Adegbola Isaac
- Michael Abolarinwa
- Olufemi Olayanju AWODOYE
- Olajide Yusuf ADEBAYO
Institutions
- Osun State University (NG)
- Federal University of Health Sciences Azare (NG)
- Federal University of Agriculture and Development Studies, Iragbiji (NG)
- Ladoke Akintola University of Technology (NG)
Publication Details
- Journal
- Romanian Cyber Security Journal
- Published
- 2026-11-12
- DOI
- https://doi.org/10.54851/v8i2y202605
- Primary Topic
- Network Security and Intrusion Detection
- Type
- article
- Field-Weighted Citation Impact
- 0.00