Securing Public Cloud Networks with Efficient Role-based Micro-Segmentation

Learn about a novel approach to securing public cloud networks through efficient role-based micro-segmentation in this 17-minute conference presentation from NSDI '25. Discover how researchers from Microsoft, Rice University, and University of Maryland tackle the critical challenge of securing network traffic within data centers as public clouds grow increasingly complex and large-scale. Explore the limitations of traditional micro-segmentation approaches that struggle with dynamic deployments and policy creation at scale. Understand how the presented system processes vast volumes of network-flow logs to accurately infer network endpoint roles by integrating domain knowledge with communication patterns in a principled manner. Examine real-world deployment evaluations demonstrating superior role inference accuracy compared to existing algorithms, and learn how this end-to-end solution achieves up to 21.5× better cost-efficiency than Apache Flink for stream processing workloads.