Fog computing enhances IoT architectures by extending resources to the network edge, yet its decentralized nature remains highly susceptible to Distributed Denial of Service (DDoS) attacks. To address these vulnerabilities, this study proposes an efficient mitigation scheme tailored for Software-Defined Network (SDN)-based Fog environments. The framework introduces a novel Anomaly-Fast Fourier Transform Auto-Regression Integrated Moving Average (FFARIMA) model to detect anomalous traffic. By converting data into the frequency domain, the model identifies rapid traffic bursts and establishes baseline behavioural patterns more effectively than traditional time-series analysis. For attack classification, the scheme employs a hybrid approach combining Random Forest (RF) and Logistic Regression (LR), leveraging their complementary strengths in pattern recognition and predictive modelling. To ensure high performance within resource-constrained fog nodes, data efficiency is optimized using Pearson’s Correlation as a feature selection technique to eliminate redundant information. This allows for faster decision-making and real-time analysis at the network edge. Performance evaluations were conducted through Google Colab using the CICDDoS2019 dataset. Results demonstrate that the proposed model significantly outperforms existing benchmarks, including Multi-level DDoS and Hypergraph clustering models, across key simulation metrics. The findings suggest that integrating frequency-based detection with hybrid machine learning provides a superior defence mechanism for large-scale IoT deployments within fog infrastructures, ensuring robust service availability against modern DDoS attacks.
Madiba et al. (2026) studied this question.