The Gcore Radar 2024 report states, the number of DDoS attacks has increased by 46% in 12 months. Supervised and unsupervised techniques struggle detecting DDoS attacks due to scarcity of labeled attack samples and an overwhelming presence of benign traffic. In contrast PU learning offers a promising solution by dividing the data into positive and unlabeled data. This study explores the effectiveness of PU learning in detecting DDoS attacks by comparing it with supervised and unsupervised methods. This method employs PU bagging, two step method and autoencoder based models to extract meaningful patterns from network traffic data, utilizing CICDDoS2017 dataset for evaluation. Proposed approach aims to improve generalizability and robustness of DDoS detection system, practically into real world scenarios where labeled data is scare. Experimental results demonstrate PU learning can achieve competitive detection accuracy while reducing reliance on label negative samples. Our findings suggest PU learning enhances adaptability to evolving attack patterns, can be integrated into existing security infrastructure for more efficient DDoS mitigation.