NeTS: Large: Measuring and Modeling Internet Choke Points as Threats to Online Freedom

点击翻译按钮获取中文摘要

Pressing policy questions with respect to Internet chokepoint, such as questions about Internet balkanization or the effects that a specific policy might have, are difficult to answer for two reasons. First, the data available are typically limited to a single country or small cross-section of the Internet for a short period of time. Second, the data do not lend themselves well to simple analysis such as averages or direct comparisons. Research is needed that bridges the gap between collecting long-term, wide-scale data and answering the ever-evolving high level questions that policy makers have. This project develops advanced capabilities for measuring Internet chokepoints and analyzing and modeling the data in a way that informs real policy questions. This project is carrying out Internet chokepoint measurements longitudinally and on a global scale, in multiple layers of the network stack including both TCP/IP layer filtering and filtering in social media. The project also studies the role of tunnels and network virtualization in how different countries and organizations implement chokepoints in order to have central locations for implementing policy. In order to accomplish these goals, the project is developing advanced TCP/IP side channel techniques for measuring TCP/IP layer filtering and understanding network topology, novel inference algorithms and natural language processing techniques to discern subtle manipulations of social media, and robust modeling techniques based on Agent Based modeling to analyze the data in a meaningful way. In terms of broader impacts, the project will provide new capabilities: (a) to inform policy makers about violations or threats to digital speech freedom, and (b) to advise democratic and anti-chokepoint efforts on how to circumvent chokepoints and limit retributions to citizens around the world. All source code developed as part of this project will be released under a GNU Public License (GPL) license, and the datasets will be made publicly available. A key part of the proposed project is mentorship of graduate and undergraduate researchers in all aspects of research, and recruitment and outreach to middle- and high-school students in New Mexico. The University of New Mexico is both a Carnegie RU/VH Very High Research Activity University and a HACU Hispanic Serving Institution as well as an EPSCoR state. This puts UNM in a unique position to involve groups that are underrepresented in computer science, and truly disadvantaged economically, in world-class Internet measurement research.