Emergency Directive 19-01 put registrar logins in the agency inventory

Emergency Directive 19-01 answered a DNS hijacking campaign that began with stolen account credentials. Its password and multi-factor orders reached registrar and DNS-operator accounts, so the inventory that mattered listed logins held at outside companies.

Looking Back revisits past weeks one at a time, up to the most recent, and asks what it meant for the banks, health systems and public agencies we serve. The series was written in 2026, with the benefit of hindsight, and each article says which week it looks back on.

Two-minute summaryThe dek and the first sentence of each section, quoted from the article

Emergency Directive 19-01 answered a DNS hijacking campaign that began with stolen account credentials. Its password and multi-factor orders reached registrar and DNS-operator accounts, so the inventory that mattered listed logins held at outside companies.

  1. The week On January 22, 2019, the Department of Homeland Security’s Cybersecurity and Infrastructure Security Agency (CISA) issued Emergency Directive 19-01, “Mitigate DNS Infrastructure Tampering”, to federal agencies.
  2. The attack path ran through logins, not agency networks The directive’s background section describes the technique in three steps.
  3. Four kinds of account, one inventory Actions Two and Three apply to “all accounts on systems that can make changes to your agency’s DNS records.”
  4. Where the directive drew its boundary A directive is defined as much by what it leaves out as by what it requires.
  5. Detection moved to a public log Actions One to Three were about prevention and cleanup.
  6. What the directive left unasked ED 19-01 was written for speed.
  7. Carrying the directive outside federal agencies ED 19-01 bound federal civilian agencies only.
  8. With hindsight CISA closed the directive on January 8, 2026.

The week

On January 22, 2019, the Department of Homeland Security’s Cybersecurity and Infrastructure Security Agency (CISA) issued Emergency Directive 19-01, “Mitigate DNS Infrastructure Tampering”, to federal agencies. The directive cited section 3553(h) of title 44 of the U.S. Code, which authorizes the Secretary of Homeland Security to issue emergency directives in response to a known or reasonably suspected information security threat, vulnerability or incident, and section 2205(3) of the Homeland Security Act of 2002, which delegates that authority to CISA’s Director. It stated that federal agencies are required to comply, and that such directives do not apply to statutorily defined national security systems or to systems operated by the Department of Defense or the Intelligence Community.

An alert came before it. The National Cybersecurity and Communications Integration Center (NCCIC), part of CISA, had published an alert on a global Domain Name System (DNS) infrastructure hijacking campaign; the alert page shows it was last revised on January 11, 2019. It said that, using compromised credentials, an attacker can change where an organization’s domain names resolve, redirect user traffic to infrastructure the attacker controls and obtain valid encryption certificates for the organization’s domain names. It pointed administrators to blogs by FireEye and Cisco Talos and recommended three practices: multifactor authentication on domain registrar accounts or other systems used to modify DNS records; verifying that DNS infrastructure points to the correct addresses or hostnames; and searching for certificates related to the organization’s domains and revoking any that were fraudulently requested.

The directive turned those recommendations into required actions. Its background section said CISA was tracking a series of incidents involving DNS infrastructure tampering, was aware of multiple executive branch agency domains that had been affected, and had notified the agencies that maintain them. It described the technique in three steps. The attacker compromises, or otherwise obtains, the credentials of an account that can make changes to DNS records. The attacker then alters records such as Address (A), Mail Exchanger (MX) or Name Server (NS) records, replacing the legitimate address of a service with one the attacker controls, so that traffic can be inspected or manipulated before it is passed on. Because the attacker can set DNS values, it can also obtain valid certificates for the organization’s domain names, which lets the redirected traffic be decrypted. In the directive’s words, “Since the certificate is valid for the domain, end users receive no error warnings.”

The directive listed four required actions:

  • Action One, audit DNS records: within 10 business days, for all .gov or other agency-managed domains, audit public DNS records on all authoritative and secondary DNS servers to verify that they resolve to the intended location, and report any that do not to CISA.
  • Action Two, change DNS account passwords: within 10 business days, update the passwords for all accounts on systems that can make changes to the agency’s DNS records.
  • Action Three, add multi-factor authentication to DNS accounts: within 10 business days, implement multi-factor authentication for all such accounts, or tell CISA which systems cannot use it, why, and when it could be enabled.
  • Action Four, monitor Certificate Transparency logs: within 10 business days, CISA would begin regular delivery of newly added certificates in Certificate Transparency logs for agency domains through its Cyber Hygiene service; on receipt, agencies were to begin monitoring immediately for certificates they did not request, and report any confirmed unauthorized certificate to the issuing certificate authority and to CISA.

A footnote defined the accounts covered by the password and authentication actions: accounts on agency-managed DNS server software, systems that manage that software, third-party DNS operators’ administration panels, and DNS registrar accounts, excluding the .gov registrar. The directive recommended password managers, recommended additional factors that are resilient to phishing, and said that, consistent with NIST SP 800-63b, SMS-based multi-factor authentication is not recommended.

Agencies were to submit a status report on January 25, 2019 and a completion report for all actions on February 5, 2019. Beginning February 6, 2019, the CISA Director would engage the chief information officers or senior agency officials for risk management of agencies that had not completed the required actions, and by February 8, 2019 CISA would report agency status and outstanding issues to the Secretary of Homeland Security and the Director of the Office of Management and Budget. The directive was to remain in effect until replaced by a subsequent binding operational directive or terminated through other appropriate action.

On January 24, 2019, CISA’s Director, Christopher Krebs, explained the directive in a post on CISA’s blog. He wrote that it was the first emergency directive CISA had issued under authorities granted by Congress in the Cybersecurity Act of 2015. He attributed the public account of the campaign to FireEye researchers, with related reporting by Cisco Talos in late 2018, and wrote that the actors had obtained access to accounts that controlled DNS records. He said the mitigations were crafted in consultation with agency IT security teams, the Office of Management and Budget and the National Institute of Standards and Technology; that in several cases the actions were basic good practices agencies may already have taken; and that monitoring Certificate Transparency may be new for some agencies. He added that, although the directive applied only to federal civilian executive branch agencies, its guidance included steps any organization could take to prevent DNS infrastructure tampering.

Emergency Directive 19-01 can be read as a DNS hygiene order: check the records, reset the passwords, add a second factor, watch the certificate logs. Read that way, it belongs with the network team, and its four actions look like tasks for the people who run name servers. We read it differently. The directive’s own description of the attack starts with a login, and its footnote on scope reaches into the web panels of registrars and DNS hosting companies. The control that mattered in January 2019 was not a server setting. It was the set of accounts, inside and outside the agency, that could change where the agency’s names pointed.

Our argument is that ED 19-01 moved DNS from a network configuration problem to an account and provider problem, and that the inventory an organization needs is a list of those accounts and of the companies that hold them. The analysis has five parts, and a closing note for organizations outside the directive’s reach:

  • The path ran through logins: what the directive says the attacker did, and why each step happened outside the agency network.
  • Four kinds of account, one inventory: the accounts the directive’s footnote names, and what an agency had to know before it could act on them.
  • Where the boundary fell: what the directive and its published questions and answers excluded, and the exposure those exclusions leave.
  • Detection in a public log: why the certificate action is a detection control that depends on knowing who asked for what.
  • What the directive left unasked: the look-back it declined to require and the controls its text does not mention.

The parts are separate. The first is about the attack, the second about the accounts, the third about the edges of the mandate, the fourth about detection and the fifth about the gaps. Together they cover what the directive required, what it assumed and what it did not do.

The attack path ran through logins, not agency networks

The directive’s background section describes the technique in three steps. None of them, as the directive describes them, involves breaking into an agency’s servers.

Step one: a credential

The first step is that the attacker compromises, or obtains by other means, the credentials of an account that can make changes to DNS records. The directive does not say where those credentials came from, and we do not either. What matters for an architect is the kind of asset involved. The asset is an account. Depending on how the agency runs its DNS, that account may live on agency-managed DNS server software, on a system that manages that software, in a third-party DNS operator’s administration panel or at a domain registrar. Two of those four places are outside the agency’s network.

The NCCIC alert framed the problem in the same terms. It said the attacker works “using compromised credentials” and put multifactor authentication on registrar accounts at the top of its recommendations. CISA’s Director, writing on January 24, said the actors obtained access to accounts that controlled DNS records. Across all three documents, the entry point is an account, and the first defensive step is to make that account harder to use.

Step two: a record change

The second step is that the attacker alters DNS records. The directive names three types: Address (A), Mail Exchanger (MX) and Name Server (NS) records. An A record gives the address for a host name, an MX record names the servers that accept mail for a domain, and an NS record names the servers that answer for the domain. The directive recommended that agencies prioritize NS records and those for key services such as mission-critical websites and mail, which fits the reach of each.

A change made through a registrar’s or DNS operator’s web panel does not need to cross the agency’s network to take effect. The agency’s firewalls, endpoint tools and internal logs would see nothing of it, because the change is made on the provider’s system, not the agency’s. The change happens in the provider’s system, and the effect shows up only when users and other systems look up the name. From the agency’s side, the first sign may be traffic going somewhere it should not, and that traffic belongs to users who are outside the agency too: members of the public, partners and remote staff.

The directive also notes that the attacker can pass the traffic on to the legitimate service once it has been inspected or manipulated. If the service keeps working, nobody reports an outage. The directive says this “creates a risk that persists beyond the period of traffic redirection”, and the Director’s post lists what an attacker in this position can do: intercepting and manipulating traffic, making services unavailable or slow, and harvesting information such as credentials or emails. Credentials captured that way can remain useful after the records are set right again, which is why the persistence point matters.

Step three: a valid certificate

The third step closes the gap that encryption would otherwise leave. Because the attacker can set DNS values for the domain, the directive says, it can obtain valid encryption certificates for the organization’s domain names, decrypt the redirected traffic and expose any data users submit. Users see no warning, because the certificate is valid for the domain.

This step turns a DNS problem into an identity problem for the whole organization. Encryption in transit is a control that a security team can point to in an audit. Here it worked as designed and still protected nothing, because the attacker had become, as far as the certificate system could tell, the legitimate holder of the name. The control that failed was further upstream: the account that let someone else speak for the domain.

Why the network team could not own it alone

The three steps cross three roles: whoever holds the registrar and hosting accounts, whoever runs the services the names point to, and whoever requests certificates. The directive’s actions follow the same split. Passwords and multi-factor authentication go to the account holders, the record audit goes to whoever can say where each name should point, and certificate monitoring goes to whoever knows which certificates were requested. A network team can run the audit. It cannot vouch for an account it does not hold or a certificate it did not request.

What the three steps add up to

Taken together, the steps say that the integrity of an organization’s DNS is only as strong as the weakest account that can change it. That is an architectural statement, not an operational one. It means DNS belongs in the same category as administrative access to an identity provider or a cloud console: a small number of accounts whose compromise lets an outsider impersonate the organization to the world. The directive did not use that language. Its actions, as the next part shows, assumed it.

Four kinds of account, one inventory

Actions Two and Three apply to “all accounts on systems that can make changes to your agency’s DNS records.” The footnote attached to both actions spells out what that means. It names four kinds of account:

  1. Accounts on agency-managed DNS server software.
  2. Accounts on systems that manage that software.
  3. Accounts on third-party DNS operators’ administration panels.
  4. DNS registrar accounts, excluding the .gov registrar.

The first two are inside the agency, on systems it runs. The last two are held at outside companies, on systems the agency does not run and cannot see into. The directive gave all four the same treatment: new passwords and multi-factor authentication within 10 business days.

An inventory the directive implied

The directive did not say “build an inventory of accounts.” It did not need to. An agency could not change the password on every account that can modify its DNS records unless it first knew every such account. That requires several layers of knowledge, each depending on the one before:

  • Every domain the agency manages, .gov and any other. Action One covers “all .gov or other agency-managed domains”, and the questions and answers on CISA’s page for the directive (they appear on the page today; the archived copy of January 23, 2019 cited here does not include them) note that an agency may have non-.gov domains registered.
  • For each domain, the registrar that holds the registration and the operator that hosts the authoritative DNS, which may be the agency itself or a hosting provider.
  • For each registrar and operator, every account that can make changes, including accounts held by contractors or set up by program offices.
  • For each account, the person or team who holds it and the authentication it uses.

A network inventory built from devices and addresses would not produce this list. It records relationships: which company holds which name, and which person can log in to change it. The difference matters for ownership. If a registrar account was opened by a program office or a contractor for a single website, the agency’s central IT staff may have no record of it. The directive’s actions made that account the CIO’s problem, with a reporting date attached.

Accounts already covered, and accounts with other protection

The questions and answers on CISA’s page for the directive narrowed the work in two places. If an agency already enforced multi-factor authentication on accounts that could change DNS records before the directive, it did not need to update the passwords on those accounts. And where an administrator had console access to a system that could make DNS changes, no action was required as long as local access was protected by other controls, such as physical ones.

Both answers point the same way. The directive cared about remote, credential-based access to the power to change records. An account that already demanded a second factor, or a console behind a locked door, was a different risk from a password-only login to a web panel reachable from anywhere.

When a provider cannot meet the bar

The directive anticipated that some systems could not support multi-factor authentication. Action Three required agencies in that position to give CISA the names of those systems, the reason, and when it could be enabled. CISA committed to review those submissions and contact agencies as needed.

The published answers went further. For a DNS hosting provider that does not offer multi-factor authentication, CISA recommended using a provider that does. For a non-.gov domain at a registrar that does not offer it, CISA wrote: “CISA recommends that you move the domain to a registrar that offers MFA.” The same answer reminded agencies that OMB memorandum M-17-06 requires executive branch agencies to use a .gov or .mil address for their public-facing digital services, with an exception for an agency that is a user or customer of a third-party site or service on a non-government domain.

That answer is where the architectural shift is plainest. A choice of registrar or DNS host had been a procurement or convenience decision. Under the directive’s questions and answers, a provider’s authentication features became a reason to leave it. The account a provider offers is part of the security architecture of every customer that relies on it.

Who answered for the accounts

The reporting schedule shows who the directive held responsible. Agencies reported status on January 25, 2019 and completion on February 5, 2019. From February 6, 2019, the CISA Director would engage the chief information officers or senior agency officials for risk management of agencies that had not finished, and by February 8, 2019 CISA would report agency status and outstanding issues to the Secretary of Homeland Security and the Director of OMB.

That chain runs to the agency’s senior technology and risk officials, not to the administrator of a name server. A registrar account that could not take multi-factor authentication, or a hosting provider that did not offer it, became an item those officials might have to explain upward. The account inventory stopped being a technical detail held by whoever set up a web panel and became something an agency’s leadership could be asked about by name.

For an architecture team, the reporting chain is a design input. If senior officials can be asked which accounts can change the organization’s names, the answer has to come from a maintained record, not from a search run under deadline. The directive’s timetable required that answer by the reporting dates. Where they could not, the gap was itself something to report.

The quality of the second factor

The directive did not stop at “add a factor.” It recommended factors resilient to phishing and, citing NIST SP 800-63b, said SMS-based authentication is not recommended. The published answers named PIV cards and security keys as phishing-resistant authenticators, and said that a combination of two single-factor authenticators is better than one and would be accepted under the directive.

The guidance matters because the attack started with credentials. A second factor that can be phished alongside the password narrows the gap without closing it. By naming the stronger forms, the directive set a bar for DNS accounts that applied even when the account lived at a company outside government.

Where the directive drew its boundary

A directive is defined as much by what it leaves out as by what it requires. ED 19-01 drew its boundary in four places, and each one tells an architect something about where the remaining exposure lies.

The statutory edge

ED 19-01 stated that directives under this authority do not apply to statutorily defined national security systems or to systems operated by the Department of Defense or the Intelligence Community. The Director’s post put the same limit in plain words: the directive applied only to federal civilian executive branch agencies that are not part of the Intelligence Community. That edge is set by law, not by CISA’s view of the risk, and the post invited any organization to use the directive’s steps anyway.

The .gov registrar

The footnote excluded the .gov registrar from the account actions. CISA’s published answer explained why: user accounts at the .gov registrar were already required to use strong passwords and two-step verification. In the answer’s words, they were excluded “not because they don’t apply, but because they are already complete.”

The exclusion is a useful precedent. It treats a provider’s built-in controls as satisfying a requirement, so long as the controls are mandatory for every account on the platform rather than optional. A registrar that offers multi-factor authentication as a setting leaves each customer to switch it on. A registrar that requires it for everyone removes that task from every customer’s list. The difference is between a feature and a property of the platform.

The provider’s own names

The published answers addressed cloud and software-as-a-service providers directly. The .gov top-level domain offered no managed DNS service, so each agency was responsible for its own name service, run itself (locally or on an infrastructure-as-a-service platform) or bought as a hosted service. The directive’s scope was agency-managed DNS infrastructure. Where an agency pointed a .gov host name at a provider’s service, the audit action covered the agency’s own name, not the provider’s domain name or infrastructure.

CISA added that agencies might choose to ask their providers how they operate their DNS securely and whether change control is protected by multi-factor authentication, but said this was not required and that providers might not share the answer.

That answer marks the residual exposure. An agency could audit its own records, reset its own passwords and enforce strong authentication on every account it held, and still depend on a provider’s DNS that it could not inspect. If the provider’s own name were redirected, the agency’s correctly configured record would still send users to the provider’s name, and from there to wherever that name now resolved. The directive’s boundary followed what the agency controlled, which is a sensible line for a mandate with a short deadline. It left the provider’s side to questions and contracts.

The past

The fourth boundary was in time. The published answers said that performing historical analysis of past record changes may be prudent but is not part of the directive. The audit action was meant to stop any current hijacking, “not a requirement to check available DNS or system logs in an effort to evaluate whether a hijack has ever occurred.” The final part of this article returns to what that choice leaves undone.

What the boundaries share

Three of the four boundaries follow control. The directive required action where the agency held the account or managed the name, excused accounts whose platform already enforced the control, and stopped at the provider’s own infrastructure. The fourth followed urgency: fix the present first. A mandate with a short deadline has to draw lines like these. An organization that is not working to a 10-business-day deadline can draw them wider, and the provider line is the first one worth moving.

Detection moved to a public log

Actions One to Three were about prevention and cleanup. Action Four was about detection, and the place it looked was outside every agency’s network.

How the action worked

Certificate Transparency logs are public records of certificates issued for domain names. The directive committed CISA to begin, within 10 business days, regular delivery of newly added certificates for agency domains through its Cyber Hygiene service. On receipt, agencies were to begin monitoring immediately for certificates they had not requested, and to report any confirmed unauthorized certificate to the issuing certificate authority and to CISA. CISA’s published answers say Cyber Hygiene reports are delivered to agencies weekly, and that agencies could also use free or paid monitoring services or build their own.

The logic follows from step three of the attack. If the attacker controls the name long enough to obtain a certificate, that certificate may appear in a public log, and the action relied on that. Watching the log for certificates nobody inside asked for gives an agency a way to notice a redirection it could not see on its own network. The Director’s post called Certificate Transparency a recent contribution from the internet security community and said monitoring it may be new for some agencies.

Choosing a monitoring source

CISA’s published answers listed three sources for the monitoring data: CISA’s own Cyber Hygiene service, free or paid monitoring services, or a monitoring setup the agency builds itself. Whatever the source, the answers said an agency would need a comprehensive list of all the second-level domains it manages, .gov and any others. A monitoring feed scoped to the domains an agency remembers will miss certificates issued for the domains it has forgotten, so the domain list from the previous part is the input to this one.

Two things the agency had to know

The published answers set out what monitoring requires. To take full advantage of it, an agency must have a comprehensive inventory of the domains it manages, and it must be able to tell whether a given certificate request was actually authorized by the organization. The second requirement is the harder one. A certificate log tells you that a certificate exists. Only the organization can say whether someone inside asked for it.

CISA described the problem in large organizations with several operating divisions, where obtaining a certificate may not be centrally managed and no single team may know that a certificate was requested. It offered two approaches: require that all certificate requests be approved by a central authority, which then monitors the log, or have the teams that run services monitor the log for the names they manage. CISA wrote: “We strongly recommend the second approach.” Where there is a single central feed, it said, the organization will need a way to share the relevant data with the teams that can verify a request, automated as far as possible.

Why the recommendation fits the argument

CISA’s preference for distributed monitoring is consistent with the account argument. The people best placed to say whether a certificate was expected are the people who run the service and hold its accounts. A central security team that sees an unfamiliar certificate has to find its owner before it can decide anything. If the inventory from the previous part exists, finding that owner is a lookup. If it does not, every alert starts with a search.

That makes the certificate request process part of the architecture too. An organization that cannot say who requested each of its certificates cannot tell a legitimate certificate from an attacker’s, and the public log stops working as a signal. The directive did not require central approval of certificate requests. It did require agencies to act on certificates they had not requested, which presumes they could tell.

What the log cannot show

Certificate monitoring catches one kind of evidence: a certificate issued for the organization’s name. A redirection that does not involve a new certificate would not appear in the log. A change to an MX record that sends mail elsewhere, for example, need not produce a certificate at all. The audit in Action One covered that case at one point in time, by checking that records resolved to the intended location. The directive did not require continuous monitoring of the records themselves, a gap the final part takes up.

What the directive left unasked

ED 19-01 was written for speed. Its required actions were near-term, in the directive’s own word, and its reporting dates ran from January 25 to February 8, 2019. A directive built that way leaves things out on purpose. Three omissions are worth naming, because an organization outside the federal mandate, or one revisiting its DNS controls now, can choose to fill them.

The look-back

The directive did not require agencies to find out whether they had already been hijacked. The published answers were explicit: the audit was to stop any current hijacking, and checking DNS or system logs for past hijacks was not required, though historical analysis may be prudent.

Set that beside the directive’s own warning that the risk “persists beyond the period of traffic redirection”. If an attacker redirected mail or a sign-in page for a period and then restored the records, an audit of current records would find nothing, while the credentials or messages captured in that window could still be in use. An organization that wants to know whether that happened needs history: the record-change logs its registrar and DNS host keep, the certificate log entries for its names over the relevant period, and its own sign-in records for accounts whose users may have been redirected. Those logs sit with providers, which brings the account inventory back into play. An organization that knows which providers hold its names knows whom to ask.

Continuous watching of records

Action One was an audit with a deadline of 10 business days. Action Four, certificate monitoring, was ongoing. Nothing in the directive’s text required agencies to watch their DNS records for changes on a continuing basis, or to receive notice when a record changed. The directive’s text also does not mention registrar or registry locks, or other provider-side controls that make record and delegation changes harder to carry out.

These are design choices for an organization to weigh, not requirements we can point to. An alert to the domain’s owners whenever an NS or MX record changes would cover the gap that certificate monitoring leaves. A provider-side lock on the registration would add a step an attacker holding a password alone could not complete. Which of these a given registrar or DNS host offers is a question for that provider, and the answer belongs in the same inventory as the accounts.

The provider’s side of the line

The third omission is the one the directive named itself: the provider’s own DNS. CISA said agencies might ask their providers how they secure their DNS and protect change control, but did not require it. For an organization that depends on outside providers for mail, websites and collaboration tools, the provider’s names carry its users as surely as its own names do.

The practical step is to ask before an incident, in due diligence and in contract terms, rather than during one. An organization can ask whether a provider enforces multi-factor authentication on the accounts that change its DNS, whether it monitors certificate logs for its names, and whether it will tell customers promptly if its names are tampered with. A provider may decline to answer, as CISA noted. A refusal is still information, and it belongs in the vendor file.

Carrying the directive outside federal agencies

ED 19-01 bound federal civilian agencies only. The Director’s post said its guidance included steps any organization could take. For a bank, a health system or a state or local agency, the directive’s value now is as a checklist written by people who had just watched the technique work against government domains.

Translate the four actions

Each action carries over without change. Audit public records for every domain the organization manages, putting name server records and key services first. Change the passwords on every account that can make changes, and add multi-factor authentication to each, preferring phishing-resistant factors over text messages. Watch certificate logs for the organization’s names, with the alerts going to the teams who can say whether a certificate was expected. Report and revoke what was not.

The directive’s order of priority carries over as well. It recommended that agencies prioritize name server records and the records for key services, giving as examples websites central to the mission, mail records and other heavily used services. For a bank, that list would start with the names customers use to sign in and the domains that send and receive mail. For a health system, it would include patient portals and the names clinical partners connect to. For a state or local agency, it would include the sites residents use to pay, apply or find election information. The ranking is ours, built on the directive’s categories, and each organization will draw it from its own services.

Add the inventory the directive implied

The harder work is the list behind the actions. For each domain, the organization needs to know the registrar, the DNS host, every account at each, who holds it and how it authenticates. That includes domains registered for campaigns, subsidiaries, acquired businesses or single projects, and accounts opened by contractors. A domain nobody remembers is a domain nobody is watching, and a registrar account held by someone who has left is an account nobody will reset.

Put the providers in scope

The directive stopped at the agency’s own names. An organization writing its own policy can go further and ask its registrars, DNS hosts and service providers the questions in the previous part. Where rules from a regulator or examiner govern how the organization oversees outside providers, it should check with counsel or its compliance team on whether those rules reach registrar and DNS accounts. We do not offer a view on that here; the point is that the accounts exist, sit at outside companies and can let someone speak for the organization’s name.

Measure whether it holds

The directive asked for completion reports, which record that actions were taken once. An organization adopting its approach can go further and keep a small set of measures that show whether the controls still hold. Useful ones include the share of domains with a named owner, a known registrar and a known DNS host; the share of accounts that can change records and use a phishing-resistant factor; the number of accounts held by people who have left or by contractors whose work has ended; and the time it takes to confirm or reject an unexpected certificate. These are our suggestions, not requirements of the directive, and each depends on the inventory existing in the first place.

Decide on the look-back

Finally, an organization should decide deliberately whether to look backward. The directive did not require it; its actions were near-term. An organization with time can ask its providers for record-change history and compare its certificate log entries against what it knows it requested. If the answer is that nothing unexpected happened, that is worth knowing too, and the request will show which providers can produce the history and which cannot.

With hindsight

CISA closed the directive on January 8, 2026. That day it announced the retirement of ten emergency directives issued between 2019 and 2024, and its release described this as the highest number of emergency directives the agency had retired at one time. ED 19-01 was first on the list. The release split the ten into two groups. Directives tied to specific vulnerabilities were retired because those vulnerabilities were by then in CISA’s Known Exploited Vulnerabilities catalog. For ED 19-01, ED 21-01 and ED 24-02, CISA said their objectives were achieved, their requirements no longer aligned with the current risk posture, and changes in practices had made them obsolete.

CISA’s page for ED 19-01 now carries the label Closed and states that CISA has concluded the directive’s intended purposes have been met, and that it has sunset the directive and all required actions it mandated. The directive had said it would remain in effect until replaced by a binding operational directive or terminated through other appropriate action. The 2026 release placed it with the directives whose objectives CISA called achieved, rather than with those tied to specific vulnerabilities now listed in the Known Exploited Vulnerabilities catalog.

The page still carries the text of the directive and its questions and answers, including the footnote naming the four kinds of DNS account and the answers about moving to providers that offer multi-factor authentication. The sources cited here do not report how many agencies completed the required actions on schedule, which agency domains were affected, or what the status and completion reports showed. We do not guess.

A separate change came to the .gov registrar itself. On March 8, 2021, CISA announced that it would begin overseeing the .gov top-level domain in April 2021, taking it over from the U.S. General Services Administration. The release said the DOTGOV Act of 2020 shifted responsibility for managing .gov to CISA. It described .gov as one of the six original top-level domains and quoted a CISA official listing among the security benefits of .gov two-factor authentication on the .gov registrar and notifications of DNS changes to administrators. It also said .gov is critical infrastructure for governments throughout the country.

The registry’s own site describes where that led. The .gov registry says it is part of CISA and manages the .gov registrar. It says it uses multi-factor authentication for all accounts in its domain registration and management system, preloads all new domains so that browsers must use HTTPS, publishes the complete list of .gov domains and recommends security practices to domain holders. Access to the registrar is managed through Login.gov. The registry says that if a person had a .gov account before November 2023, the old credentials cannot be used with the new registrar, and that first-time domain requesters must verify their identity through Login.gov. It recommends adding two authentication methods to a Login.gov account.

Those changes cover one of the four kinds of account in the directive’s footnote, the registrar account, for .gov names. The registrar account was the one the directive had already excluded because its controls were complete, and it is now reached through Login.gov, with identity verification for first-time requesters. The other three kinds are where they were. The registry says it does not offer DNS hosting and does not host .gov websites or email. An agency, county or city that uses .gov still runs its own DNS or buys it from a provider, and the accounts that change those records are still its own to inventory and protect.

That is how the directive’s argument looks with hindsight. The government kept strengthening the account it controls, at the registry, where strong authentication was already a property of the platform rather than a setting each holder had to choose. The accounts at DNS hosts and service providers remain where ED 19-01 found them, outside the agency’s network and inside its attack surface. In our view, the directive’s footnote is still the right starting list for any organization that wants to know who can change where its names point.

Spotted an error? Tell us and we will check it and correct the article where needed.

  • Emergency Directive 19-01
  • DNS
  • CISA
  • multi-factor authentication
  • certificate transparency
  • .gov