With the wide application of cloud computing and 5G technology, modern call platforms have evolved from traditional telephone switching systems to intelligent service platforms based on IP networks. However, the open network architecture also makes these platforms become the main target of network attacks, especially Distributed Denial-of-service (DDoS) attacks presenting new features such as diversified attack methods, continuously expanding scale, and more precise targets. In this paper, an intelligent dynamic defense system based on a generative adversarial network (GAN) is proposed to train a detection model with strong generalization ability by simulating diverse attack scenarios. The system adopts a closed-loop architecture of “detection-defense-optimization”, which contains three core modules: Conditional GAN (CGAN) -based attack traffic generation module, improved MobileNetV3 edge detection module, and dynamic resource elastic scheduling module. Experimental results show that the system maintains 98.5% detection accuracy in normal business traffic scenarios, and the false alarm rate is controlled within 0.5%. Compared with the traditional threshold detection system, the method in this paper improves the detection rate of new attacks by about 20%, and at the same time, the response delay of the system is controlled within 10ms, which fully meets the real-time requirements of the call platform. This study provides a new solution for the security protection of call platform, and shows excellent performance in three dimensions: attack detection, resource scheduling and log analysis.