in this guide
Every account your company has, from the bank portal to the payroll system, eventually trusts one thing: an email arriving at an address on your domain. Whoever controls where that domain points controls where those emails land. That control lives in two or three places most companies never look at after the day they were set up: the registrar account, the DNS host, and the registrar’s support desk.
The founder departure guide asks whose name is on those accounts when a person leaves. This guide asks the other question: who, today, could repoint your domain without you noticing, and what sits in their way.
Your domain is the master key to every password reset you have.
The recovery address matters here: who else can read the invoices inbox covers role addresses, and the domain hijacking topic has the numbers.
what moving a domain actually means
A hijacker rarely needs to steal the domain in the legal sense. Changing where it points is enough, and that is a setting.
CISA described the pattern in Emergency Directive 19-01 of 22 January 2019, written after a wave of DNS tampering against US government domains. The attacker gets the credentials of an account that can change DNS records. They change the address, mail or nameserver records to point at their own servers. Because they now answer for the domain, they can obtain valid certificates for it, so the fake site shows a padlock and no warnings. The directive is now closed. The pattern it describes matches the cases below.
what one nameserver change hands over
- 1a login to the registrar or DNS hostStolen, reset, or handed over by support.
- 2nameservers or records changedThe domain now answers from the attacker's servers.
- 3website and mail followVisitors reach a copy of your site. Email to your addresses reaches the attacker.
- 4password resets follow mailEvery service that trusts an address on your domain can be reset from outside.
- 5certificates follow the domainDomain validation passes, so the copy looks legitimate.
The Security Alliance (SEAL) made step four explicit in an advisory of 21 August 2025: change the mail records and you receive the domain’s email, and with it the password resets for accounts tied to it. Its advice after a recovery is to look for admin accounts the attacker created while they held the domain.
the four doors
Four routes lead to the same nameserver change. Most companies have thought about one of them.
who can repoint your domain
- the registrar loginthe account you think of, usually with one admin
- the registrar's support deskcan reset that login for a convincing caller
- the DNS hostoften a different vendor, a different login, and API tokens
- the recovery inboxoften an address on the same domain it protects
rarely reviewedusually known
The registrar login. The account where the domain was bought. It can change nameservers, contacts and locks, and it can request the code that moves the domain to another registrar. At many small companies it has one user, sometimes a personal account, and whatever MFA was offered the year it was opened. The founder departure guide shows what happens when that one user leaves.
The registrar’s support desk. Every registrar has a way back in for customers who lost their password. That process is an identity control run by someone else, under someone else’s rules. It is the same weakness covered in the helpdesk reset guide, except you cannot train this help desk or read its scripts.
The DNS host. Many companies keep the domain at one vendor and the DNS records at another, such as their website host or a content delivery network. That second account can change every record, and it often has API tokens of its own. Cloudflare’s API documentation, for example, lists a zone permission called DNS Write that grants write access to DNS. A token with that permission, made for a deployment script years ago, is a door with no person behind it. The identity you can’t offboard explains why such tokens outlive the project that created them.
The recovery inbox. The email address on the registrar account is where reset links and transfer confirmations go. If it is an address on the same domain, the domain protects its own recovery. Once someone repoints the mail, the next reset link goes to them.
three cases, three doors
Most of the cases below hit cryptocurrency projects, because a hijacked front end pays the attacker within minutes. The mechanism is the same for an accounting firm or a clinic, and the payoff is your email.
A migration that changed the login (July 2024). In 2023 Google sold its domain registration business to Squarespace, and customers’ domains were moved over in stages. Between 9 and 12 July 2024, attackers took over Squarespace accounts holding migrated domains and changed their DNS. Krebs on Security counted at least a dozen organisations hit, mostly crypto projects. Squarespace’s own postmortem says a limited number of accounts were compromised through “a weakness related to OAuth logins”, fixed on 12 July, and that the migration made no changes to MFA. Researchers quoted by Krebs gave a different account: migrated domains could be claimed by anyone who signed up first with the email address tied to them, without verifying it. Krebs also wrote that multi-factor authentication had been disabled during the migration, which Squarespace denies. The two accounts have not been reconciled. Both point to the same lesson: a vendor change can change who can log in, and nobody asks you first. The full timeline is on the Squarespace domains incident page.
A support desk that reset the login (July 2024). The exchange dYdX published a postmortem on its own domain. It survived the first takeover on 9 July because DNSSEC settings left in place at the registrar stopped the changed nameservers from resolving. On 23 July the attacker came back another way. According to dYdX, Squarespace’s customer service performed a “human-initiated reset”, removed two-factor authentication, changed the account’s email to the attacker’s address and did not contact the other admins listed on the account. The attacker removed DNSSEC this time. dYdX says two users lost about $31,000, and it moved the domain to another registrar the next day. MFA on the account did nothing, because the support desk could take it off. MGM Resorts is the same move against an in-house help desk.
A registrar that was the weak point (May 2025). On 12 May 2025 the curve.fi domain was repointed at the registrar level, according to Curve’s incident post. The attacker changed the domain’s nameserver delegation to their own server. Curve says it saw no sign of compromised login credentials, describes the registrar’s response as “slow and uncooperative”, and moved its users to a new domain. Its post said the method of compromise was still under investigation. Sometimes the door is inside the registrar, and the only control you have is which registrar you chose and what lock you asked for.
An older case shows how quiet this can be. The write-up of the perl.com hijacking, published on 28 February 2021, says the domain was compromised in September 2020, moved to another registrar in December, and noticed only on 27 January 2021, when the domain started resolving to different addresses. The author speculates the registrar was social-engineered with fake documents. One of his lessons: keep at least one contact address off the domain, because when the domain goes, mail to it goes too.
what the locks actually stop
Registrars offer a “registrar lock”, and many companies take it as the answer. It covers less than the name suggests.
The registrar lock blocks transfers out. ICANN’s page on domain status codes describes clientTransferProhibited as telling the registry to reject requests to transfer the domain. Client codes are set by the registrar, which means anyone logged into your registrar account can turn them off. In the Squarespace and dYdX cases nobody needed a transfer: the attackers changed nameservers inside the account, and a transfer lock never comes into play.
The transfer code moves it to another registrar. ICANN’s registrant FAQ says an Auth-Code is required to move a domain between registrars, and your registrar must provide it within five calendar days of a request. Whoever holds the registrar login can usually generate it in the dashboard.
The registry lock adds a phone call. Above the registrar sits the registry that runs the top-level domain. ICANN’s status page notes that some registry operators offer a Registry Lock Service, and server codes set by the registry take precedence over the registrar’s. Verisign, which runs .com and .net, describes its version on its own registrar resources page: it covers .com, .net, .cc and .name, it is offered through participating registrars, and unlocking needs an authorised person at the registrar to receive a call from Verisign and give an individual security phrase. A stolen password or a persuasive support call does not get past that. It adds delay to legitimate changes and is offered only by some registrars and registries, so ask yours.
DNSSEC can turn a quiet takeover into a loud one. In the dYdX case it kept the first set of changed nameservers from resolving. The second time, the attacker held the account and switched it off. It is worth having, and it protects nothing once someone controls the account where it is set.
the settings that close the doors
Closing these doors takes an afternoon and someone who owns the result.
- Name an owner. One person accountable for the registrar and the DNS host, and a second person who can get in. Write both down where the IT provider can find them.
- Make the recovery address independent. The email on the registrar and DNS accounts should be a role address that does not depend on the domain it protects, or at least one contact should sit outside it. The Security Alliance advisory recommends a separate address for the registrar account for this reason.
- Use the strongest MFA the registrar offers. CISA’s directive asked for MFA on every account that can change DNS and advised against text messages. Security keys or an authenticator app, on every user of the account.
- Ask the registrar how its support desk resets an account. What does a caller have to prove, and does the registrar contact the other admins before removing MFA? If yours cannot answer, that is the answer.
- Count the DNS tokens. List the API tokens and users on the DNS host. Delete the ones nobody can name.
- Ask about registry lock for the one domain your email runs on. That domain carries every password reset you have.
- Watch for changes. Nameserver and mail record changes should alert someone. CISA’s directive also asked agencies to watch Certificate Transparency logs, the public record of every certificate issued, for certificates they did not request. The certificates guide covers who can issue certificates for your domain.
If a hijack does happen, the first call is to the registrar, and your break-glass email address must still work. Find out now which address that is. The break-glass guide covers keeping an emergency path that does not run through the thing that failed.
When the registrar replies, read it as an attacker would. What would you need to know about your company to pass their check?
on the record
all myths →sources
- CISA: Emergency Directive 19-01, Mitigate DNS Infrastructure Tampering (22 January 2019)
- Squarespace status page: Domain Hijacking postmortem (July 2024)
- Krebs on Security: Researchers: Weak Security Defaults Enabled Squarespace Domains Hijacks (15 July 2024)
- dYdX: DNS Nameserver Hijacking Postmortem (July 2024)
- Curve Finance: Curve Domain Incident (May 2025)
- Security Alliance (SEAL): advisory on domain hijackings (21 August 2025)
- perl.com: The Hijacking of Perl.com (28 February 2021)
- ICANN: EPP Status Codes, what do they mean and why should I know?
- ICANN: FAQs for registrants, transferring your domain name
- Verisign: Registry Lock
- Cloudflare Developers: API token permissions
