| Popis: |
In an SDN environment, topology discovery is a key function that allows the network controller to accurately understand how network devices are connected and interact. Traditional topology discovery methods typically rely on specific protocols, such as LLDP (Link Layer Discovery Protocol), but these methods may have issues like inefficiency and security vulnerabilities. Therefore, researchers and engineers have been exploring more secure and efficient topology discovery methods to adapt to the evolving network demands and challenges. This paper introduces a novel SDN topology discovery method, Attopo, based on the attention mechanism and network flow analysis, independent of traditional protocols. We apply the attention mechanism to SDN topology discovery for the first time and address the security vulnerabilities of the OFDP protocol through a non-protocol mechanism. Attopo combines Self-attention, Cross-attention, and Prob-attention mechanisms to analyze network flows received by the controller and extract topology information, thereby enhancing the model’s accuracy and efficiency in network topology discovery. Additionally, we have collected a dataset containing network flows and topology labels by simulating different network environments, which enhances the model’s adaptability. Experimental validation shows that our method performs well in terms of accuracy, recall rate, F1 score, and AUC metrics in topology discovery. |
