dnstracker: Measuring Centralization of DNS Infrastructure in the Wild

Abstract

The Internet Domain Naming System (DNS) is one of the pillars for the Internet and has been the subject of various Distributed Denial-of-Service (DDoS) attacks over the years. As a countermeasure, the DNS infrastructure has been engineered with a series of replication measures, such as relying on multiple authoritative name servers and using IP anycast. Even though these measures have been in place, we have seen that, when servers rely on third-party DNS providers for reliable services, there may be certain levels of infrastructure centralization. In this case, an attack against a DNS target might affect other authoritative DNS servers sharing part of the infrastructure with the intended victim. However, measuring such levels of infrastructure sharing is a daunting task, given that researchers typically do not have access to DNS provider internals. In this paper, we introduce a methodology and associated tool dnstracker that allows measuring, to various degrees, the level of both concentration and shared infrastructure using active DNS measurements. As a case study, we analyze the authoritative name servers of all domains of the Alexa Top 1 Million most visited websites. Our results show that, in some cases, up to 12.000 authoritative name servers share the same underlying infrastructure of a third-party DNS provider. As such, in the event of an attack, those authoritative DNS servers have increased the probability of suffering from collateral damage.