DDOS Attacks Detection via Deep Learning Approaches

Abstract

Abstract: The proliferation of Distributed Denial of Service (DDoS) attacks presents a pressing challenge to network security. Conventional rule-based detection methods are increasingly inadequate against the evolving tactics employed by cyber adversaries. This journal proposes a novel approach integrating Support Vector Machines (SVM) into advanced machine learning architectures for fortified DDoS detection. The research methodology initiates with comprehensive data collection, gathering diverse network traffic scenarios and DDoS attack instances. This dataset becomes the foundation for subsequent phases, employing sophisticated feature engineering to extract vital patterns for model development. The feature selection process involves using feature engineering techniques including Data Collection, Model Development and SVM Integration, to extract the most discriminative and relevant features from the network traffic data. The primary innovation lies in the creation of a hybrid CNN-LSTM model, capable of discerning spatial and temporal dependencies within network data, thereby augmenting DDoS threat identification. The integration of SVM into machine learning and deep learning paradigms forms the crux of this study. By leveraging SVM's classification proficiency alongside the CNN-LSTM model's capabilities, a sophisticated DDoS detection system aims to surpass conventional rule-based limitations.