Generative Autonomous Systems Topology Generator

Accurate modeling of Internet topologies is crucial for evaluating communication protocols and traffic control strategies. However, existing topology generators often rely on rigid, structure-oriented assumptions - such as scale-free models - and fail to capture key structural properties, including betweenness centrality, clustering, and assortativity. Moreover, most current tools offer limited usability and adaptability, with outdated implementations and minimal user interfaces. In this article, we present a novel software tool, Generator of Internet Autonomous Systems Topology (Gen-AS), which implements the Deep-generative graphs for the Internet (DGGI) model - a datadriven generator for Autonomous System (AS) graphs. DGGI learns topological patterns directly from data using deep learning, enabling it to replicate complex network properties without relying on predefined structural assumptions. Designed to be flexible and user-friendly, Gen- AS features a Command Line Interface (CLI) and a Graphic User Interface (GUI), making it accessible to a wide range of users. Experimental evaluations show that Gen-AS significantly outperforms existing topology generators, achieving average improvements of 93.22% in assortativity, 94.43% in betweenness centrality, 98.87% in clustering, and 95.09% in node degree distributions. These results, measured using the Maximum Mean Discrepancy (MMD) metric, highlight Gen-AS's effectiveness in producing realistic Internet topologies.