Distributed denial-of-service (DDoS) attacks causes significant threats in modern cloud and edge computing because of complex traffic patterns that intricate attack detection. Traditional methods have limitations in distinguishing attack strategies and extracting relevant information. This study retains machine learning and deep learning techniques to find DDoS attacks using the CICDDoS-2019 dataset, which has pre-processed to include 431,371 records and 78 features. The preprocessing intricate handling duplicates, missing values, and class imbalances using SMOTE. Various classifiers, including Logistic Regression, Decision Tree, Random Forest, XGBoost, and LightGBM, alongside One-Class SVM and Autoencoder-Bidirectional LSTM (AE–BiLSTM) models, were utilized. Assessments of metrics such as accuracy, precision, recall, F1-score, and ROC–AUC were employed to assess the models. Results indicated that ensemble methods like XGBoost and LightGBM achieved perfect ROC–AUC scores of 1.00, while AE–BiLSTM shows strong anomaly detection with a ROC–AUC of 0.9979. This framework gives a practical solution for improving network security within cloud and edge computing environments.