TopoIntent: Automated Compliance-Checked Network Topologies

Arxiv pdf 2026-08-01T00:00:00
arXiv Paper — PDF not available. Only the Executive Summary is available here. To read or download the full paper, visit the arXiv abstract page.

Abstract

Enterprise security topology design begins before device configuration. Architects must translate business intent, regulatory requirements, and risk assumptions into zones, boundary devices, inter-zone paths, and access-control policies. Existing NetOps automation tools mostly operate after this design has been fixed: they translate policies into configurations or interpret existing diagrams, but provide limited support for producing a structured security topology from an underspecified natural-language request. We describe TopoIntent, a system that compiles security intent into executable, compliance-checked network topologies. TopoIntent uses a schema contract to constrain topology generation, retrieves reference architectures from a curated template library through dense-vector search, and applies staged fusion to separate intent-template alignment from IG-level security completion. It then checks the generated topology against CIS Controls v8.1.2 safeguards whose evidence is visible at the topology layer, records cases requiring manual review, and repairs structural gaps through additive schema-preserving edits. The final topology is exported to Mininet scripts with kernellevel iptables ACLs, so that reachability and allow/deny behavior can be exercised rather than inspected only as a static diagram. We build an evaluation set from reference security architecture diagrams because no public benchmark exists for this requirement-to-topology task. The retrieval reference set contains 22 templates and 44 synthetic intents across five scenarios; the held-out evaluation set contains 7 templates and 14 intents from finance and government scenarios excluded from the retrieval index. We evaluate TopoIntent with structural coverage, topology-visible CIS satisfaction, Mininet policy tests, and diagnostic feedback from Stage 5 to Stage 4. On the heldout set, additive repair raises first-pass topology-visible CIS satisfaction from 0.78/0.78 (IG2/IG3) to 1.00 in fewer than 1.5 rounds on average. In the end-to-end ablation, one feedback round raises the post-ACL policy pass rate from 0.78 to 0.88.

Loading executive summary...

LINK COPIED TO CLIPBOARD