Implementing Machine Learning (ML) Based Hybrid Model to Mitigate Distributed Denial of Service (DDoS) attacks in Multi-Access Edge Computing (MEC) Environment

The fifth-generation mobile network (5G-MN) technology is regarded as an evolution of traditional networks originated from the first, second, third, and fourth generation mobile network respectively. All generational mobile networks are vulnerable to security concerns that are common to traditional mobile communication networks. The security concerns can be hackers compromising the networks by sending malicious packets to authorize user devices which eventually lead to breach of legitimate users’ data and also causes high latency in the networks. This security concern evolves with the evolution of communication networks. Therefore, the main objective of this study is to implement a robust mitigation scheme in Mobile Edge Computing (MEC) due to the limitations of traditional approaches in addressing evolving security concerns. MEC is one of the significant technologies in the implementation of 5G-MN. MEC enhances the capabilities of cloud computing and brings computational capability closer to the network edge. This reduces latency in 5G, resulting in better end-user experience. This study focuses on MEC security concerns called distributed denial of service (DDoS) attacks which can occur at different network layers of the network. The aim was to tackle security issues such as DNS flooding attacks and identify effective techniques for mitigating DDoS attacks in MEC systems. We suggested employing Supervised Machine Learning (SML) algorithms as the proposed solution or mitigation techniques, which are Random Forest (RF), Decision Tree (DT), Naïve Bayes (NB), K-Nearest Neighbor (K-NN), Logistics Regression (LR), and Blending/Stacking Model (BM/SM). Subsequently, we then conducted an evaluation of these techniques. The evaluation was based on multiple performance metrics such as accuracy, detection/recall, precision, F1-Measure, Matthew’s correlation coefficient (MCC). Machine learning (ML) models were stacked to implement ML-based hybrid model. Our evaluation demonstrated that hybrid models surpassed traditional ML models in performance, with RF being the most effective in mitigating DDoS attacks in MEC. The results were supported by statistical analysis such as Probability Density Function (PDF), Area Under the Receiver Operating Characteristic (AUROC) and hypotheses testing, reinforcing the superiority of hybrid models.