With the development of satellite communications, the number of satellite nodesis constantly increasing, which undoubtedly increases the difficulty of maintaining network security. Combining software defined network (SDN) with traditional space-based networks provides a new class of ideas for solving this problem. However, because of the highly cen?tralized network management of the SDN controller, once the SDN controller is destroyed by network attacks, the network it manages will be paralyzed due to loss of control. One of the main security threats to SDN controllers is Distributed Denial of Service (DDoS) attacks, so how to detect DDoS attacks scientifically has become a hot topic among SDN security man?agement. This paper proposes a DDoS attack detection method for space-based networks based on SDN architecture. This attack detection method combines the optimized Long Short-Term Memory (LSTM) deep learning model and Support Vector Machine (SVM), which can not only make classification judgments on the time series, but also achieve the purpose of detecting and judging through the flow characteristics of a period of time. In addi?tion, it can reduce the detection time as well as the system burden.