Firefly-Driven Feature Selection for Enhanced DDoS Detection in IoT Networks

Internet of Things (IoT) applications are experiencing a significant surge in popularity. However, this rapid growth is accompanied by various challenges, for example, the gigantic volume of information delivered, network versatility issues, and security issues. Since most detection systems circumvent security, distributed denial of service (DDoS) assaults are common in IoT devices. Due to the notable variations in signatures and traffic rate, in the recent literature, a multitude of models have been described to identify them, but the problem continues to be difficult to solve. This work presents a new detection technique by diminishing the feature space, which brings down compu-tational intricacy and overfitting using the Firefly algorithm with Correlation-Based Weighting with K-Nearest Neighbors (K - NN) as the underlying classifier. To evaluate the efficacy of our proposed architecture, Our investigations focused on the CICDDoS-2019 Day 1 dataset. The experimental findings demonstrate that our proposed model outperforms other methods such as K-Nearest Neighbors (KNNs), Logistic Regression, and LightGBM (LGBM) classifier. Specifically, our model exhibits superior performance compared to these traditional techniques, especially when paired with the RandomForest (RF) classifier.