Optimized Cyber Attack Detection in IoT Networks Using Feature Selection and LightGBM
Akshat Gaurav, Brij Bhooshan Gupta, Kwok Tai Chui · 2024
The spread of IoT networks presents major security issues, especially with regard to identifying and reducing cyberattacks. In this context, we proposed an optimal LightGBM model for IoT network cyber attack detection. We used Random Forest classifier to find optimal features. Then LightGBM is trained on the selected features. After this the LightGBM model was tested against eleven conventional machine learning models. With the greatest accuracy ($\mathbf{9 9. 8 \%}$), F1 score $(\mathbf{9 9. 8 \%}$), precision ($\mathbf{9 9. 8 \%}$), and recall $\mathbf{(9 9. 8 \%}$), the LightGBM model clearly outperformed other machine learning models.