- Joseph Sventek
The mobile networked systems (e.g., 3G/4G and upcoming 5G) have offered ubiquitous Internet services to billions of mobile devices (smartphones, tablets, IoT devices, etc.). Despite being working systems, however, they are also suffering from network failures, performance downgrades and security threats with alarming frequency. A core reason is their complexity and opaqueness: The large-scale network infrastructure suffers from the intricate protocol interactions and configuration/policy conflicts (due to ad-hoc designs and operations), while the mobile clients lack the crucial insights of the underlying "black-box" network operations to take appropriate actions.
In this talk, I will present techniques to address these limitations on both sides. First, for the network infrastructure, I will show how domain-specific formal methods can be developed to facilitate the detection and resolution of problematic protocol interactions, and configuration/policy conflicts. Second, for the mobile clients, I will show how a data-driven system design helps enhance end intelligence without changing network infrastructure. I will present MobileInsight, an open-source software tool that makes it possible by collecting, analyzing and exploiting full-stack "black-box" cellular network information over commodity phones. Both results suggest a paradigm shift in the future mobile networked system (e.g., 5G). I will end by briefly exploring this vision.
Yuanjie Li is a Ph.D. candidate in Computer Science at UCLA, advised by Professor Songwu Lu. His interests are in the areas of networked systems, mobile computing and security. In particular, he works on creating effective tools and primitives to enable fast, reliable and secure services in the state-of-art and the next-generation mobile systems. Before joining UCLA, he received his B.E. from Tsinghua EE in 2012. He is the recipient of ACM MobiCom’16 Best Community Paper Award, and UCLA Dissertation Year Fellowship in 2016.