OptCosineDCNN: An Advanced Network Attack Detection and Mitigation in SDN Environments Utilizing Enhanced Deep Learning Based Approaches
D. Mahesh, T. Sampath Kumar · Cybernetics & Systems · 2025
Software-Defined Networking (SDN) has become a popular choice for network management due to its flexibility, improved performance, and cost efficiency. However, existing SDN-based Intrusion Detection Systems (IDS) often struggle with low detection rates, reliance on predefined rules, and inefficient mitigation strategies. To address these limitations, we propose an optimized Cosine Deep Convolutional Neural Network (OptCosineDCNN) for accurate DDoS attack detection and classification. Our approach integrates Multi-Scale Residual Deep CNN (MSRD-CNN) for feature extraction and the Copula Entropy-based Golden Jackal Optimization (CEGJO) algorithm for feature selection, enhancing model efficiency. Additionally, hyperparameter tuning using Boosted Dung Beetle Optimization (BDBO) optimizes model performance, reducing overfitting and improving generalization. In the final analysis, this procedure produces more accurate and efficient predictive results by improving accuracy, decreasing overfitting, and generalizes effectively to new data. To lessen these threats, we have implemented a mitigation strategy. Using the four benchmark datasets, the suggested IDS model is experimentally assessed and achieves high performance in terms of accuracy and precision. Also, we ran several tests in various scenarios Free- Attack, No Mitigation-Attack, and Mitigation-Attack, to assess the efficacy and efficiency of our suggested DDoS mitigation system. These outcomes show how reliable our suggested mitigation system is at both preventing DDoS attacks and ensuring the uninterrupted operation of regular network services.