Bridging Code and Graphic: A Cross-Mode Framework for Large-Scale Network Scenario Construction

The construction of large-scale virtual network scenarios (VNS) is essential for security evaluation, architecture planning, and the validation of emerging network technologies. However, existing code-driven approaches produce lengthy and low-readability scripts, which hinder early error detection; meanwhile, graphical-driven methods lack automation and become prohibitively inefficient for large-scale VNS construction. Moreover, purely graphical inspection becomes ineffective for large, cyclic topologies, often leading users to overlook critical configuration errors - thereby incurring unnecessary deployment overhead and increased rollback costs. This paper proposes a cross-mode framework bridging code-driven automation and graphic-driven intuitiveness, which can address key challenges in large-scale VNS construction: (1) achieving consistent semantic mapping between heterogeneous Infrastructure as Code (IaC, e.g., Terraform, Pulumi) scripts and graphical topologies via a unified graph-object abstraction; (2) designing a novel hybrid MST-DFS algorithm to reduce user disorientation in cyclic topologies and ensure accurate and efficient intent verification; (3) through a mixed-method evaluation that combines quantitative testing with 24 participants and qualitative interviews, we validate the framework's effectiveness and practical stability: this approach reduces error rates by up to 92%, shortens task time by up to 90%, and enables over 85% of errors to be detected prior to deployment. Furthermore, it has been stably deployed for over five months in the PengCheng Cyber Range, supporting more than 1,000 virtual network systems across 10+ industries.