With the rapid development of big data and artificial intelligence technology, federated learning, as a new data privacy protection technology, is gradually becoming an effective means to solve the problem of data independence and realize the invisibility of data availability. At the same time, how to better promote the development of security vulnerability identification and defense mechanism in the federal learning system has gradually become the focus of managers. From the actual application of the current federated learning system, it still faces many security threats such as data leakage and model poisoning, which have a lot of impact on the security management of the current and subsequent federated learning system. Based on the research on the identification and defense mechanism of security vulnerabilities in the federated learning system, this paper aims to propose an effective identification and defense mechanism through data analysis, so as to provide a direction for further research on defense mechanisms.