Software-Defined Networking (SDN) offers centralized programmability an‎d global visibility, positioning it as a suitable platform for intelligent an‎d adaptive security solutions. Many Machines Learning (ML) an‎d Deep Learning (DL) based Anomaly Detection Systems (ADS) have been suggested, but their eva‎luation methods are still scattered, inconsistent, an‎d too dependent on old datasets an‎d limited metrics. We propose the SDN Anomaly Detection eva‎luation Framework (SADEF), a benchmarking approach intended to facilitate systematic an‎d reproducible testing of anomaly detection systems inside realistic emulated SDN settings. SADEF is different from other methods since it uses a multi-dimensional set of assessment measures, such as detection accuracy, computing overhead, network performance impact, an‎d adversarial robustness. SADEF connects theoretical developments with real-world use by combining contemporary traffic statistics, different attack scenarios, an‎d adversarial machine learning methods. The main goal of this framework is to create a stan‎dardized an‎d comprehensive benchmark that enables academics an‎d practitioners to create, compare, an‎d implement more reliable, efficient, an‎d robust ADS solutions for SDN networks.

10/29/2025 12:00:00 AM

87999

1404/08/08

Benchmarking an‎d eva‎luation of Anomaly Detection Systems in Software Defined Networking

Software-Defined Networking (SDN) , Anomaly Detection System (ADS) , Machine Learning (ML) , Deep Learning (DL) , Benchmarking Framework