The methods that are being traditionally used are unable to match the increasing threat landscape, as DDoS attacks are becoming more complex. Machine learning has improved the detection and helps in decreasing real-time attacks. This paper focuses on deep learning-based various supervised and unsupervised methods to detect the DDoS attacks. While the unsupervised learning methods, such as clustering, offer benefits in the identification of new attacks with labelled data, the supervised learning algorithms, like Support Vector Machines (SVM) and Decision Tree, are being used to classify patterns. On an extensive scale, dynamic assault scenarios, deep learning models, “Convolutional Neural Networks (CNNs) and Recurrent Neural Networks (RNNs)” had a remarkable performance. The KDDCup99 and CICIDS datasets, which are deployed to train these models, have also been examined in this review. The key performance metrics, including accuracy, precision, and recall, are used to gauge the models’ efficiency. To inform future research and improve cybersecurity defences against DDoS attacks, this study also observes the recent innovations, datasets, and performance metrics.