Within Software-defined networking (SDN) architectures, the central controller assumes the responsibility of managing the entire network. However, this network-wide perspective is acquired through link discovery service, thereby exposing SDNs to vulnerabilities like Link Fabrication Attacks (LFAs).

This research, centered on enhancing the security of SDN, particularly targets vulnerabilities related to LFAs and aims to improve topology discovery services. Initially, the study focused on investigating SDN vulnerabilities, leading to the development of the Link Latency Attack (LLA), a new method devised to evade modern defense strategies such as TopoGuard+. The effectiveness of LLA was confirmed through practical testing in environments utilizing the Floodlight controller and Mininet emulation. The second phase of the study involved the creation of the Machine Learning Link Guard (MLLG), a sophisticated ML-based detection system tailored for SDNs. MLLG was designed for real-time detection of LFAs and LLAs, integrating seamlessly with existing SDN controllers and adaptable to various network conditions. To complement MLLG, the Real-time Link Verification (RLV) system was developed for networks with frequent topological changes. Together, MLLG and RLV demonstrated their efficacy in a range of network scenarios, with MLLG excelling in more static settings and RLV adapting effectively to dynamic environments.

As telecommunications increasingly move towards softwareization, the final objective of this research became crucial. The study’s in-depth exploration into LFAs within these emerging infrastructures led to the identification of the Bearer Migration Poisoning (BMP) attack in the Open Radio Access Network (RAN), a significant and novel vulnerability. This finding emphasized the necessity for adaptable defense mechanisms in such software-centric environments. Notably, the RLV system, initially developed for traditional SDN, was successfully adapted to the OpenRAN architecture. This adaptation showcases the potential of ML-based systems like RLV to provide robust and efficient defenses, essential for securing the dynamic telecommunications infrastructure. The research profoundly influences telecom’s future, especially in advancing open, software-defined 6G. Adapting RLV for OpenRAN pioneers network security, potentially inspiring more Open RAN research and fostering telecom innovation.