AutoML for Network Security

The project creates a network-aware, autonomous attack detection system tailored for Software-Defined Networks (SDN). Existing solutions are deemed unsuitable for SDN environments because they often rely on additional infrastructure to manage high data rates and either overlook or assume static traffic patterns. To address these limitations, the proposed mechanism incorporated metrics such as varying traffic loads, diverse data rates, and detection time to avoid overfitting.

The project has two core modules: AutoFE and AutoML. The AutoFE module is responsible for selecting the optimal feature engineering approach based on the network’s current state, while the AutoML module identified the most suitable machine learning algorithm for the task. By balancing feature engineering techniques, machine learning models, and the dynamic conditions of network traffic, the project advances the field of autonomous attack detection in SDN.

Furthermore, a comprehensive software architecture is developed and executed, which incorporates a variety of machine learning algorithms and feature engineering techniques. This architecture is created using open-source network tools, which render it both practical and accessible for further research and application.