The existing wireless communication interference methods rely heavily on the characteristics of the target signals obtained in communication reconnaissance, require complex prior knowledge, and have problems in keeping up with the dynamic changes of relevant parameters. This paper proposes a GAN-based technology of generating spoofing-jamming I/Q signals. The jammer will be able to deceive and interfere with opponents through GAN by generating spoofing-jamming I/Q signals highly correlated to real I/Q signals without prior knowledge. This paper first introduces the principle of GAN and the GAN model to be adopted; uses a software radio system composed of LabView and NI USRP software and hardware plat-forms to simulate real communication scenarios to collect real I/Q signal data; generates spoofing-jamming I/Q signals through GAN model; uses t-SNE algorithm to perform dimensionality reduction on both the real and the spoofing-jamming I/Q signal data to visualize their distributions; finally, tests the spoofing-jamming I/Q signals on receiver’s pre-trained classifier. The experimental results show that GAN provides an effective method for generating high-quality spoofing-jamming I/Q signals.