Feature Selection and 1DCNN-based DDOS Detection in Software-Defined Networking

Software-defined networking (SDN) revolutionizes network management by offering centralized control over complex infrastructures, but it also introduces significant security vulnerabilities. particularly Distributed Denial of Service (DDoS) attacks that significantly interrupt network services. The challenge of efficiently detecting DDoS attacks in SDNs is exacerbated by the computational overhead associated with analyzing numerous network features using conventional Machine Learning (ML) techniques. Addressing this gap, our research proposes a novel Intrusion Detection System (IDS) utilizing a 1D Convolutional Neural Network (1DCNN-IDS) model specifically designed to identify DDoS threats within SDN environments. To refine feature selection and enhance detection accuracy, we applied a hybrid objective function incorporating the Akaike Information Criterion (AIC), F-test (ANOVA), and T-test. The effectiveness of our model was validated using three diverse datasets: InSDN, CICIDS2017, and UNSW-NB15, achieving impressive accuracies of over 98%, 96%, and 92% respectively, alongside high precision, recall, and F1 scores. These findings highlight the substantial potential of incorporating ML and Deep Learning (DL) techniques for effective and efficient intrusion detection in SDNs, highlighting our methodology's contribution towards mitigating DDoS attack risks in these networks.