This study addresses the challenges in cybersecurity education, including the complexity of knowledge structures, limitations of traditional teaching resources, and data security risks associated with existing tools, by developing an automated knowledge graph-based teaching platform. The platform was constructed using Neo4j LLM Graph Builder, with a multi-model adaptation engine designed to integrate domestic large models for localized data processing. The MITRE ATT&CK framework was incorporated as core teaching content, and three visualization modules—ECharts interactive graphs, Web-based structured tables, and native Neo4j visualization—were developed. The constructed knowledge graph contains 1,265 entities and 5,828 relationships. Empirical results demonstrate significant improvements in students' conceptual understanding and practical skills, with a 35% increase in classroom interactions, a 25% improvement in assignment completion efficiency, and overall user satisfaction exceeding 89%. This research provides an innovative solution for intelligent teaching under data security constraints.