Software-Defined Networking (SDN) offers superior programming ability, administration, adaptability, and productivity compared to traditional networks by separating the data and control planes. A single point of control over network devices is essential for mitigating the impact of Distributed Denial of Service (DDoS) attacks, as it provides a holistic view of the network and facilitates traffic analysis to recognize malicious actions. Nevertheless, even while the separation of control and data planes offers advantages, this system’s design is still vulnerable to DDoS attacks, which present difficulties in prompt identification and counteraction. An optimal set of features that can accurately detect these attacks is necessary to prevent them effectively. Our work emphasizes feature representation by utilizing the encoder component of Autoencoder to enhance DDoS attack detection on a self-generated SDN dataset. Real-time detection is achieved by simulating attack traffic on a High-Performance Computing (HPC) system. We evaluated the performance of the ONOS Flood Defender Application. We found that LightGBM classifier performs best in detecting DDoS attack with a detection time of 11.61 ms and an accuracy of 99.1%.