In a network environment that become more complex due to the continuous update of emerging Internet technologies, distributed denial of service (DDoS) attacks also become easier to implement and more difficult to detect. Such as in the blockchain environment, the performance of node equipment is often low, the deployment cost of existing detection systems is high, and multi-point collaborative detection cannot be performed. In response to these problems, this paper proposes a lightweight DDoS attack detection method based on LightBGM. The paper proposes a multi-component network traffic feature group to extract the features of network flow to realize the lightweight of DDoS attack detection on the feature input. The experimental results show that the proposed feature group can describe the network flow more accurately. Compared with other DDoS attack detection methods, LightGBM-based methods have higher accuracy, recall and prediction rates, lower error rate and missing rate, while occupying less computing resources.