A Multi-Layer Adaptive Security Approach for Overcoming Security Challenges in Software-Defined and Data Center Networks

Abstract

Software-Defined Networks (SDNs) and Data Center Networks (DCNs) are becoming fundamental components in cloud computing and other large-scale digital services. SDNs and DCNs pose new security challenges due to their centralized control plane, network virtualization, dynamic orchestration of resources, and other novel features. Cybersecurity risks associated with SDNs and DCNs include DDoS and other types of attacks, resource exploitation, traffic diversion, and data leakage. Most other types of traditional intrusion detection systems do not adapt or provide real-time protection for high-scale programmable networks. Our research proposes a multi-layer adaptive security approach that fuses Dynamic Threat Detection (DTD), Adaptive Access Control (AAC), Secure Network Virtualization (SNV), Behavior-based Anomaly Detection (BAD), and Policy-Driven Security (PSF). The security framework created using fusion proposes deep learning-based anomaly detection, entropic (entropy) access control, fusion of virtualization protection over homomorphic encryption, and command-based dynamic orchestration as protective mechanisms to secure multiple levels of one or more networks. This new approach achieved a detection rate of 97. 8%, precision as 97. 2%, recall as 96. 9%, and an F1 score of 97. 0 with a ROC-AUC of 0.987. The model achieved detection latency (6.4 ms) under 20 Gbps throughput with strong scalability and high bandwidth. The proposed framework shows improved detection performance, fewer false positives, and better resilience for the network in comparison to CNN-LSTM, Transformer-based, and federated learning intrusion detection systems. Results prove the combination of adaptive intelligence, secure virtualization, and dynamic policy enforcement boosts cybersecurity defenses in unique ways for programmable SDN and DCN infrastructures.