Fuzz testing automatically uncovers vulnerabilities in software. However, existing fuzz testing tools for network protocols are not able to fully explore their internal state space, resulting in limited coverage. Finite state machines comprehensively model the implementation of network protocols to provide an in-depth understanding of their system behavior and internal state space. This study proposes a fuzz testing method for network protocols based on finite state machines. It focuses on the commonly used TLS protocol, using finite state machine learning to model the implementation of the TLS protocol, reflecting the protocol’s internal state space and system behavior. Subsequently, guided by finite state machines, the fuzz testing of the TLS protocol achieves deeper depth and broader code coverage. This study also implements a prototype system, SNETFuzzer, which outperforms existing methods in important metrics such as coverage in a series of comparative experiments. SNETFuzzer successfully discovers multiple vulnerabilities, including two new ones, demonstrating its practicality and effectiveness.