A recent series of blog posts analysed the centralization of the Internet. These posts cover the diversity of authoritative name servers, the diversity of MX records, an analysis of the use of CAA records across generic Top-Level Domains (gTLDs) and finally A, AAAA and CNAME records.
However, different domains are not necessarily of the same importance, especially to users.
One use case of domains is to simply ‘park’ them, thus the domain is not used for any relevant service such as mail or a webpage, but only a simple website is hosted, displaying advertisements or sales banners. This approach is used to monetize registered but unused domains and is merchandised by large parking services.
My fellow researchers from the Technical University of Munich (TUM), the University of Twente and I set out to identify parking services and quantify the impact of domain parking on the DNS ecosystem and evaluate their impact on the Internet’s centralization.
Domain parking services are offered by different operators, for example, GoDaddy, Sedo, or Parking Crew. These operators host websites on parked domains including advertisements and sales banners, and manage visibility and cash flows. Domain owners can register their domains with these services and park their assets using the DNS in one of several ways.
- Domains can be parked by delegating authority for a domain name to the nameservers of a parking service. Under this approach, the name server delegation (NS records) will point to the service-specific name servers.
- Domain name owners can use their own name servers but configure an IP address record (A or AAAA) or canonical name record (CNAME) and point it to the infrastructure of the parking service.
Both approaches can be inferred from DNS data, for example, by actively querying for records. We systematically collected a list of 82 parking services. Depending on the specific parking service, different types of indicators allow for parked domains to be identified. For example, GoDaddy (Free Parking) uses specific IP addresses (22.214.171.124 and 126.96.36.199), while AfterNic relies on a set of nameservers (ns*.afternic.com). We published the complete list of services and all indicators to allow the identification of parking services on all DNS datasets.
We conduct DNS scans from TUM, targeting more than 334M domain names.
This measurement is seeded with domains from:
- Well over 1k of available zone files from the Centralized Zone Data Service (CZDS), which includes legacy gTLDs and newer gTLDs.
- Names on the Cloudflare Radar, Majestic, and Umbrella Top 1M lists.
- A static collection of 98.1M domains from 52 ccTLDs (partial zones, for example, 22M .tk and 13 M .de domains).
Our identification shows that 58.5M (17.5 %) out of 334M domains are parked. This value is only a lower bound as we might miss smaller parking services and do not consider additional techniques to park pages, for example, based on HTTP redirects. Figure 2 shows the share of parked domains for different sources. Domain parking is no phenomenon of new generic TLDs such as .live but also drastically impacts legacy gTLDs (.com, .net, and .org) with up to 30% of parked domains. Furthermore, even top lists contain parked domains.
Table 1 provides an overview of the top five parking services. GoDaddy has a market share of more than 50% if both the free and CashParking service is combined. They acquired several other parking services throughout the years and included them in their own portfolio. HugeDomains, the second largest provider, focuses only on nine eTLDs including .com and .net.
|GoDaddy (Free Parking)||28.69M||572|
To further illustrate the impact of domain parking on DNS data and studies related to the centralization of the Internet, we analyse their impact on hosting providers. We identify the hosting provider of a domain based on the A and AAAA records, mapping the IP addresses to their respective provider.
Table 2 shows the top five organizations hosting parked domains. Their rank is based on the number of hosted domains. While AS396982 operated by Google is the second most important Autonomous System Number (ASN) based on hosted domains, 90% of these domains are only parked and thus mostly insignificant for users.
If we do not consider parked domains, the ASN is only on rank 13. The IP address (188.8.131.52) hosting most domains (compare A, AAAA, and CNAME records) is operated by Google but explicitly used by GoDaddy for their free parking service. AS40034, Confluence Networks, is impacted even more with 99.9% of parked domains. Excluding parked domains, its rank drops from 21 to only 4,344.
|ASN||Organization||Rank||Domains||Domains||Ratio||Rank w/o parking|
|14618||Amazon AES||5||7.77M||3.32M||42.8 %||8|
|29873||Newfold||19||2.1 M||0.64M||30.2 %||20|
Our findings show that most parked domains are concentrated on a few services, often relying on single hosting locations. Furthermore, because parking services largely focus on advertisements and domain sales, the content of parked pages is only circumstantially relevant for Internet users.
Domain parking is of low importance for most users and the Internet in general but represents a large part of the DNS and web ecosystem, and can introduce bias in analyses that treat parked domains the same as any other. Major shares of domains hosted within, for example, Google Cloud or Amazon AWS are only parked domains.
Based on their specific appearance and client value, they need to be classified as a specific asset and evaluated as such.
You can check out our paper ‘Domain Parking: Largely Present, Rarely Considered!‘ to read an in-depth analysis. We published our list of parking services and DNS indicators on GitHub. This set of indicators can be used to easily filter parked domains from DNS data or can be applied by OpenINTEL during collaborations.
This post is based on an evaluation of new data and differences to the original paper are visible, for example, GoDaddy’s CashParking currently uses Amazon to host parked domains instead of Google as reported in 2022.
Johannes Zirngibl is a Research Associate at TU Munich. His research interests include analysis and deployment of network protocols in the Internet, DNS deployment and centralization, and scalable network scans.
The views expressed by the authors of this blog are their own and do not necessarily reflect the views of APNIC. Please note a Code of Conduct applies to this blog.