See also my Google citations.


Assessing Support for DNS-over-TCP in the Wild

Jiarun Mao, Michael Rabinovich, and Kyle Schomp
In Proceedings of the 2022 Passive and Active Measurement Conference (PAM ’22), Virtual, March 2022

While the DNS protocol encompasses both UDP and TCP as its underlying transport, UDP is commonly used in practice. At the same time, increasingly large DNS responses and concerns over amplification denial of service attacks have heightened interest in conducting DNS interactions over TCP. This paper surveys the support for DNS-over-TCP in the deployed DNS infrastructure from several angles. First, we assess resolvers responsible for over 66.2% of the external DNS queries that arrive at a major content delivery network (CDN). We find that 2.7% to 4.8% of the resolvers, contributing around 1.1% to 4.4% of all queries arriving at the CDN from the resolvers we study, do not properly fallback to TCP when instructed by authoritative DNS servers. Should a content provider decide to employ TCP-fallback as the means of switching to DNS-over-TCP, it faces the corresponding loss of its customers. Second, we assess authoritative DNS servers (ADNS) for over 10M domains and many CDNs and find some ADNS, serving some popular websites and a number of CDNs, that do not support DNS-over-TCP. These ADNS would deny service to (RFC-compliant) resolvers that choose to switch to TCP-only interactions. Third, we study the TCP connection reuse behavior of DNS actors and describe a race condition in TCP connection reuse by DNS actors that may become a significant issue should DNS-over-TCP and other TCP-based DNS protocols, such as DNS-over-TLS, become widely used.

Partitioning the Internet using Anycast Catchments

Kyle Schomp and Rami Al-Dalky
In ACM SIGCOMM Computer Communication Review (CCR ’20), October 2020

In anycast deployments, knowing how traffic will be distributed among the locations is challenging. In this paper, we propose a technique for partitioning the Internet using passive measurements of existing anycast deployments such that all IP addresses within a partition are routed to the same location for an arbitrary anycast deployment. One IP address per partition may then represent the entire partition in subsequent measurements of specific anycast deployments. We implement a practical version of our technique and apply it to production traffic from an anycast authoritative DNS service of a major CDN and demonstrate that the resulting partitions have low error even up to 2 weeks after they are generated.

Akamai DNS: Providing Authoritative Answers to the World's Queries (slides)

Kyle Schomp, Onkar Bhardwaj, Eymen Kurdoglu, Mashooq Muhaimen, and Ramesh K. Sitaraman
In Proceedings of the 2020 ACM SIGCOMM Conference (SIGCOMM ’20), Virtual Event, NY, USA August 2020

We present Akamai DNS, one of the largest authoritative DNS infrastructures in the world, that supports the Akamai content delivery network (CDN) as well as authoritative DNS hosting and DNS-based load balancing services for many enterprises. As the starting point for a significant fraction of the world's Internet interactions, Akamai DNS serves millions of queries each second and must be resilient to avoid disrupting myriad online services, scalable to meet the ever increasing volume of DNS queries, performant to prevent user-perceivable performance degradation, and reconfigurable to react quickly to shifts in network conditions and attacks. We outline the design principles and architecture used to achieve Akamai DNS's goals, relating the design choices to the system workload and quantifying the effectiveness of those designs. Further, we convey insights from operating the production system that are of value to the broader research community.

A Look at the ECS Behavior of DNS Resolvers

Rami Al-Dalky, Michael Rabinovich, and Kyle Schomp
In Proceedings of the 2019 Internet Measurement Conference (IMC ’19), Amsterdam, Netherlands, October 2019

Content delivery networks (CDNs) commonly use DNS to map end-users to the best edge servers. A recently proposed EDNS0-Client-Subnet (ECS) extension allows recursive resolvers to include end-user subnet information in DNS queries, so that authoritative DNS servers, especially those belonging to CDNs, could use this information to improve user mapping. In this paper, we study the ECS behavior of ECS-enabled recursive resolvers from the perspectives of the opposite sides of a DNS interaction, the authoritative DNS servers of a major CDN and a busy DNS resolution service. We find a range of erroneous (i.e., deviating from the protocol specification) and detrimental (even if compliant) behaviors that may unnecessarily erode client privacy, reduce the effectiveness of DNS caching, diminish ECS benefits, and in some cases turn ECS from facilitator into an obstacle to authoritative DNS servers’ ability to optimize user-to-edge-server mappings.

Characterization of Collaborative Resolution in Recursive DNS Resolvers (slides)

Rami Al-Dalky and Kyle Schomp
In Proceedings of the 2018 Passive and Active Measurement Conference (PAM ’18), Berlin, Germany, March 2018

Recursive resolvers in the Domain Name System play a critical role in not only DNS' primary function of mapping hostnames to IP addresses but also in the load balancing and performance of many Internet systems. Prior art has observed the existence of complex recursive resolver structures where multiple recursive resolvers collaborate in a "pool". Yet, we know little about the structure and behavior of pools. In this paper, we present a characterization and classification of resolver pools. We observe that pools are frequently disperse in IP space, and some are even disperse geographically. Many pools include dual-stack resolvers and we identify methods for associating the IPv4 and IPv6 addresses. Further, the pools exhibit a wide range of behaviors from uniformly balancing load among the resolvers within the pool to proportional distributions per resolver.

Towards a Model of DNS Client Behavior

Kyle Schomp, Michael Rabinovich, and Mark Allman
In Proceedings of the 2016 Passive and Active Measurement Conference (PAM ’16), Heraklion, Crete, Greece, March 2016

The Domain Name System (DNS) is a critical component of the Internet infrastructure as it maps human-readable hostnames into the IP addresses the network uses to route traffic. Yet, the DNS behavior of individual clients is not well understood. In this paper, we present a characterization of DNS clients with an eye towards developing an analytical model of client interaction with the larger DNS ecosystem. While this is initial work and we do not arrive at a DNS workload model, we highlight a variety of behaviors and characteristics that enhance our mental models of how DNS operates and move us towards an analytical model of client-side DNS operation.

Is The Web HTTP/2 Yet?

Matteo Varvello, Kyle Schomp, David Naylor, Jeremy Blackburn, Alessandro Finamore, and Konstantina Papagiannaki
In Proceedings of the 2016 Passive and Active Measurement Conference (PAM ’16), Heraklion, Crete, Greece, March 2016

Version 2 of the Hypertext Transfer Protocol (HTTP/2) was finalized in May 2015 as RFC 7540. It addresses well-known problems with HTTP/1.1 (e.g., head of line blocking and redundant headers) and introduces new features (e.g., server push and content priority). Though HTTP/2 is designed to be the future of the web, it remains unclear whether the web will---or should---hop on board. To shed light on this question, we built a measurement platform that monitors HTTP/2 adoption and performance across the Alexa top 1 million websites on a daily basis. Our system is live and up-to-date results can be viewed at isthewebhttp2yet.com. In this paper, we report findings from an 11 month measurement campaign (November 2014 - October 2015). As of October 2015, we find 68,000 websites reporting HTTP/2 support, of which about 10,000 actually serve content with it. Unsurprisingly, popular sites are quicker to adopt HTTP/2 and 31% of the Alexa top 100 already support it. For the most part, websites do not change as they move from HTTP/1.1 to HTTP/2; current web development practices like inlining and domain sharding are still present. Contrary to previous results, we find that these practices make HTTP/2 more resilient to losses and jitter. In all, we find that 80% of websites supporting HTTP/2 experience a decrease in page load time compared with HTTP/1.1 and the decrease grows in mobile networks.

multi-context TLS (mcTLS): Enabling Secure In-Network Functionality in TLS

David Naylor, Kyle Schomp, Matteo Varvello, Ilias Leontiadis, Jeremy Blackburn, Diego Lopez, Konstantina Papagiannaki, Pablo Rodriguez, and Peter Steenkiste
In Proceedings of the 2015 ACM SIGCOMM Conference (SIGCOMM ’15), London, England, August 2015

Transport Layer Security (TLS), is the de facto protocol supporting secure HTTP (HTTPS), and is being discussed as the default transport protocol for HTTP2.0. It has seen wide adoption and is currently carrying a significant fraction of the overall HTTP traffic (Facebook, Google and Twitter use it by default). However, TLS makes the fundamental assumption that all functionality resides solely at the endpoints, and is thus unable to utilize the many in-network services that optimize network resource usage, improve user experience, and protect clients and servers from security threats. Re-introducing such in-network functionality into secure TLS sessions today is done through hacks, in many cases weakening overall security.

In this paper we introduce multi-context TLS (mcTLS) which enhances TLS by allowing middleboxes to be fully supported participants in TLS sessions. mcTLS breaks the "all-or-nothing" security model by allowing endpoints and content providers to explicitly introduce middleboxes in secure end-to-end sessions, while deciding whether they should have read or write access, and to which specific parts of the content. mcTLS enables transparency and control for both clients and servers.

We evaluate a prototype mcTLS implementation in both controlled and "live" experiments, showing that the benefits offered have minimal overhead.More importantly, we show that mcTLS can be incrementally deployed and requires small changes to clients, servers, and middleboxes, for a large number of use cases.

DNS Resolvers Considered Harmful (slides)

Kyle Schomp, Mark Allman, and Michael Rabinovich
In Proceedings of the 2014 ACM SIGCOMM Workshop on Hot Topics in Networks (HotNets ’14), Los Angeles, CA, October 2014

The Domain Name System (DNS) is a critical component of the Internet infrastructure that has many security vulnerabilities. In particular, shared DNS resolvers are a notorious security weak spot in the system. We propose an unorthodox approach for tackling vulnerabilities in shared DNS resolvers: removing shared DNS resolvers entirely and leaving recursive resolution to the clients. We show that the two primary costs of this approach—loss of performance and an increase in system load—are modest and therefore conclude that this approach is beneficial for strengthening the DNS by reducing the attack surface.

Assessing DNS Vulnerability to Record Injection (slides) (data)

Kyle Schomp, Tom Callahan, Michael Rabinovich, and Mark Allman
In Proceedings of the 2014 Passive and Active Measurement Conference (PAM ’14), Los Angeles, CA, March 2014

The Domain Name System (DNS) is a critical component of the Internet infrastructure as it maps human-readable names to IP addresses. Injecting fraudulent mappings allows an attacker to divert users from intended destinations to those of an attacker's choosing. In this paper, we measure the Internet's vulnerability to DNS record injection attacks-including a new attack we uncover. We find that record injection vulnerabilities are fairly common-even years after some of them were first uncovered.

On Measuring the Client-Side DNS Infrastructure (slides) (data)

Kyle Schomp, Tom Callahan, Michael Rabinovich, and Mark Allman
In Proceedings of the 2013 Internet Measurement Conference (IMC ’13), Barcelona, Spain, October 2013

The Domain Name System (DNS) is a critical component of the Internet infrastructure. It allows users to interact with Web sites using human-readable names and provides a foundation for transparent client request distribution among servers in Web platforms, such as content delivery networks. In this paper, we present methodologies for efficiently discovering the complex client-side DNS infrastructure. We further develop measurement techniques for isolating the behavior of the distinct actors in the infrastructure. Using these strategies, we study various aspects of the client-side DNS infrastructure and its behavior with respect to caching, both in aggregate and separately for different actors.

Complexity and Security of the Domain Name System

Kyle Schomp
PhD Dissertation, Case Western Reserve University, Cleveland, OH, 2016

The Domain Name System (DNS) provides mapping of meaningful names to arbitrary data for applications and services on the Internet. Since its original design, the system has grown in complexity and our understanding of the system has lagged behind. In this dissertation, we perform measurement studies of the DNS infrastructure demonstrating the complexity of the system and showing that different parts of the infrastructure exhibit varying behaviors, some being violations of the DNS specification. The DNS also has known weaknesses to attack and we reinforce this by uncovering a new vulnerability against one component of the system. As a result, understanding and maintaining the DNS is increasingly hard. In response to these issues, we propose a modification to the DNS that simplifies the resolution path and reduces the attack surface. We observe that the potential costs of this modification can be managed and discuss ways that the cost may be mitigated.

Dynamic TCP Proxies: Coping with Mobility and Disadvantaged Hosts in MANETs

Kyle Schomp
Master Thesis, Case Western Reserve University, Cleveland, OH, 2010

TCP proxies have been introduced as a method to improve throughput and reduce congestion in mobile ad hoc networks. Proxies split the path into several shorter paths which have higher throughput due to reduced packet loss and round trip time. As a side effect, congestion is reduced because fewer link layer retransmissions occur. In current protocols, proxies are assigned at the start of the transfer and must be used for the duration. Due to mobility and congestion change, pinned proxies can actually reduce throughput. In this thesis, we present a second version of the DTCP protocol which includes the ability to switch proxies in the middle of a transfer. We demonstrate in the Network Simulator version 2 that the new protocol performs better than other related protocols in simulated mobile ad hoc networks with varying levels of mobility and congestion.