Testing the robustness and security of network protocol implementationsis essential across all domains. We present Network-Fuzzer, a generic, feedback-driven network fuzzer designed to test andanalyze protocol implementations by operating directly on real traffic.Unlike traditional code coverage-based fuzzers, NetworkFuzzer worksat the network level and employs a closed-loop fuzzing mechanism thatdynamically adapts based on server responses. The system incorporatesthree key components: (i) response-aware fuzzing operators that performprotocol-specific packet mutations, (ii) a Conditional Tabular GAN (CTGAN)model that learns from both normal and abnormal traffic to generatediverse and protocol-compliant test cases, and (iii) Large LanguageModels (LLMs) that automate the generation of testing rules from protocolspecifications. While NetworkFuzzer is protocol-agnostic andapplicable to a wide range of network protocols, in this paper we focuson its application to the Digital Imaging and Communications inMedicine (DICOM) protocol, which is commonly used for medical imageexchange, to demonstrate its utility in healthcare cybersecurity. Ourevaluation shows that NetworkFuzzer effectively executes real-worldattacks and generates realistic synthetic traffic, thus enhancing the robustnessof testing and training for security systems.