UC San Diego

Computer networks play a central role in the transmission of information across theworld. Autonomous systems (administrative domains or “ASes”) are the building blocks of such wide-area networks and are responsible for delivering traffic to their individual subscribers as well as to other networks to which they are connected. So-called transit ASes—who sell access to the rest of the Internet to customer ASes for a fee—are mostly invisible to end users but may be able to operate on their traffic, for instance by observing unencrypted traffic or metadata, or by tampering with specific network flows serving popular applications. In many countries, transit ASes serve as the principal intermediaries between domestic access ASes and the global Internet.

In this dissertation, we introduce the concept of transit influence, which quantifies theexposure of an AS or groups thereof, such as in an industrial sector, or nation, to observation and tampering by a specific transit network. We hypothesize that there are countries where transit agreements are the dominant form of international connectivity, and where specific ASes have significant degrees of transit influence (TI) over the country as a whole as well as over individual organizations within them. We quantify TI by developing three metrics at distinct granularities: at the country level (CTI), at the AS level (ATI), and at the sectoral level (W-ATI). In order to apply these methods, we first identify 75 countries—with approximately 1 billion Internet users, on aggregate—where transit providers are the dominant mode of inbound connectivity, using analyses of existing interconnection data and our own large-scale measurement campaign. Applying CTI, we find 32 nations that have transit ecosystems with concerning topological features: traffic destined to over 40% of their IP addresses is exposed to a single network. We further study the AS topologies of three nations in South America and find a small number of transit ASes that exert out-sized influence in several sectors, including finance, utilities, and education. We validate our findings with in-country network operators at 123 ASes in 19 countries, who confirm that our results are consistent with their understanding of their countries’ networks.