A Systematic Review of Deep Learning Models for DDoS Attack Detection in Cybersecurity

Abstract

Distributed Denial of Service (DDoS) attacks are still one of the biggest threats to cybersecurity. They overload network resources and cause problems for businesses, governments, and important infrastructure. Conventional detection systems frequently inadequately respond to the dynamic and extensive characteristics of these attacks. Deep learning (DL) has become a powerful method in recent years because it can automatically learn complex traffic patterns, adjust to new attack strategies, and cut down on false positives. This systematic literature review (SLR) evaluates the role of deep learning models in DDoS attack detection by synthesizing 97 peer-reviewed studies published from 2015 to 2024. The review looks at popular deep learning (DL) methods like Convolutional Neural Networks (CNNs), Recurrent Neural Networks (RNNs), Long Short-Term Memory (LSTM), Autoencoders, Generative Adversarial Networks (GANs), and hybrid models. It also looks at benchmark datasets (like CICIDS2017, CICDDoS2019, and NSL-KDD) and evaluation metrics like accuracy, precision, recall, F1-score, and detection rate. Key findings show that DL methods are better at detecting things than traditional methods, but they have a lot of problems, such as dataset imbalance, high computational cost, lack of real-time adaptability, and being vulnerable to attacks from other systems. The study delineates research deficiencies in the creation of lightweight, interpretable, and scalable models appropriate for IoT and edge computing environments. This review synthesizes existing knowledge, offering insights into the advantages and drawbacks of DL-based DDoS detection while delineating avenues for future research aimed at developing resilient, adaptive, and reliable cybersecurity systems.