As cloud computing becomes increasingly central to modern business infrastructure, ensuring the security of cloud-native systems is critical. Kata Containers enhances the deployment of thousands of applications across Alibaba’s platforms by providing a customized Linux kernel optimized for cloud workloads. However, traditional fuzzers are inadequate for these environments due to discrepancies between test and actual deployment scenarios. This paper introduces DragonRadar, a novel fuzzing framework adapted for Kata containers, which aligns the testing environment with cloud deployment realities. The evaluation shows that DragonRadar effectively identifies 25 kernel vulnerabilities in the mainline Linux kernel used in the Kata Containers environment, while maintaining a similar code coverage to vanilla Syzkaller. By far, DragonRadar is deployed on Alibaba’s CI/CD pipeline to ensure the security of the kernel in their cloud-native environment. DragonRadar is available at https://github.com/TOBESTONG/DragonRadar.