A Siamese-Based Approach for Network Intrusion Detection Systems in Software-Defined Networks.

Recently, a new approach to networking called Software-Defined Networking (SDN) has emerged based on the idea of separating the centralized control plane from the data plane, which simplifies network management and meets the needs of modern data centers. However, the centralized nature of SDN also introduces new security risks that could hamper widespread SDN adoption, such as single points of failure. The controller is a critical vulnerability since an attacker who compromises it can control traffic routing and severely disrupt the network. SDN is still an emerging technology, utilizing deep learning for Network Intrusion Detection Systems (NIDS) is an effective security solution that could enable more accurate and adaptive threat detection to against attacks targeting vulnerabilities introduced by centralized control. In this paper, we describe a Siamese-based method for NIDSs in SDN. When it comes to the process of training and testing models based on Siamese Networks, making effective pairs is a key strategy that can have a considerable impact on the outcome. To prevent overfitting, we enhance the data pairing both within and across classes. The findings of our methodology demonstrate a notable enhancement in the efficacy of NIDS, resulting in an accuracy rate of approximately 100%. This estimated accuracy exceeds that of baseline methods. The study’s conclusions facilitate the development of reliable IDS systems tailored for SDN environments.