BOD 18-01 set federal email and web security deadlines that DHS could verify from outside
DHS’s October 16, 2017 directive required DMARC, HTTPS-only service and retired ciphers on a dated schedule, checked by DHS scans. In our reading, it made an agency’s internet-facing configuration its compliance record, an approach later policy kept.
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
DHS’s October 16, 2017 directive required DMARC, HTTPS-only service and retired ciphers on a dated schedule, checked by DHS scans. In our reading, it made an agency’s internet-facing configuration its compliance record, an approach later policy kept.
- The week On Monday, October 16, 2017, the Department of Homeland Security issued Binding Operational Directive 18-01, Enhance Email and Web Security.
- The inventory: what an agency had to count as its own The directive’s actions applied to second-level domains, mail-sending hosts, internet-facing mail servers and publicly accessible websites and web services.
- The mail path: from STARTTLS and p=none to a reject policy in one year The email requirements came in stages.
- The web path: HTTPS everywhere, HSTS, preloading and retired ciphers The web requirements built on M-15-13.
- The evidence: external scans, DMARC reports and monthly status reports The directive used three kinds of evidence.
- How the directive reached different organizations The directive bound federal civilian executive branch agencies.
- Objections, in order The argument that the directive made internet-facing configuration the compliance record, and that this approach outlasted its settings, faces objections.
- What an agency CIO or CISO could check The directive’s specific deadlines are long past.
- With hindsight The directive’s page on CISA’s site records one adjustment to its settings.
The week
On Monday, October 16, 2017, the Department of Homeland Security issued Binding Operational Directive 18-01, Enhance Email and Web Security. A binding operational directive is a compulsory direction to federal executive branch departments and agencies to safeguard federal information and systems. DHS issues them under the Federal Information Security Modernization Act of 2014. They do not apply to national security systems or to certain Defense Department and intelligence community systems.
The directive explained its reasoning in terms of cyber hygiene. It said agencies could protect the integrity and confidentiality of internet-delivered data, minimize spam and better protect users from phishing email that appears to come from a government system. It said the requirements rested on current network scan data and a clear potential for harm.
The email section named four technologies. STARTTLS lets a receiving mail server signal that it can encrypt mail in transit. SPF and DKIM let a sending domain mark its legitimate mail. DMARC tells receivers what the domain owner wants done with mail that fails those checks. The directive said a DMARC policy of reject gives the strongest protection against spoofed email. DMARC reports can also tell an agency where forgeries come from.
The web section built on earlier policy. In 2015, OMB memorandum M-15-13 had required all existing federal websites and web services to be reachable only over HTTPS, with HTTP Strict Transport Security. In 2017, the .gov registry began automatically preloading new federal .gov domains as HSTS-only in modern browsers. The directive said agencies had to make more progress, including removing support for known-weak protocols and ciphers.
The directive gave a figure from DHS’s own scans. According to DHS’s Cyber Hygiene scanning data, seven of the ten most common vulnerabilities seen across federal agency networks when the directive was issued would be addressed by complying with its web security actions.
The required actions ran on a calendar, which the directive’s compliance guide translated into dates. Within 30 days, by November 16, 2017, each agency had to send DHS an agency plan of action. At 60 days, December 15, 2017, it had to report on progress, and then every 30 days until done.
Within 90 days, January 15, 2018, agencies had to configure all internet-facing mail servers to offer STARTTLS. All second-level domains had to have valid SPF and DMARC records. The DMARC policy had to be at least p=none, with at least one address set to receive aggregate or failure reports.
By 120 days, February 13, 2018, all publicly accessible federal websites and web services had to serve HTTPS only, with HSTS. Agencies had to identify second-level domains that could be HSTS preloaded and send DHS the list. SSLv2 and SSLv3 and the 3DES and RC4 ciphers had to be disabled on web and mail servers.
Two more actions followed. Within 15 days after the National Cybersecurity and Communications Integration Center set up a central reporting location, agencies had to add DHS as a recipient of DMARC aggregate reports. Within one year, by October 16, 2018, they had to set a DMARC policy of reject for all second-level domains and mail-sending hosts.
DHS gave itself duties too. It would review each plan, coordinate preload lists with the DotGov program, and track and verify progress through scanning by its National Cybersecurity Assessments and Technical Services team. The directive acknowledged possible budget effects. It told agency CIOs and procurement officers to coordinate with their chief financial officers.
Binding Operational Directive 18-01 is easy to read as a checklist of protocols. Its more lasting feature was how compliance would be shown. Every required setting could be observed from the public internet. DNS records show SPF and DMARC policies. Mail servers announce STARTTLS. Web servers reveal their protocols, ciphers and HSTS headers. DHS said it would rely on its own scanning to track and verify progress.
That choice moved the burden of proof. An agency could not satisfy the directive by writing a policy or certifying a plan. It had to make its public-facing systems answer correctly when anyone asked. The directive also routed DMARC reports to DHS, so the department would see who was sending mail in each agency’s name.
Our argument is that the directive made an agency’s internet-facing configuration its compliance record, and that this approach outlasted the directive’s specific settings. We take its design in four parts:
- The inventory: what an agency had to count as its own.
- The mail path: from STARTTLS and p=none to a reject policy in one year.
- The web path: HTTPS everywhere, HSTS, preloading and retired ciphers.
- The evidence: external scans, DMARC reports and monthly status reports.
The first part defines the scope. The second and third set the controls. The fourth decides how anyone knows the controls are in place. Together they cover the directive from the agency’s side. After the four parts we look at how it reached different organizations, at objections, and at what an agency CIO or CISO could check.
A note on sources and our reading. We describe the directive from the DHS text now hosted by CISA, which includes its compliance guide and later notes. We describe later events from OMB, CISA and the .gov registry. Statements about what the directive required are our reading of those documents. Agencies should rely on CISA’s current text and their own counsel.
The inventory: what an agency had to count as its own
The directive’s actions applied to second-level domains, mail-sending hosts, internet-facing mail servers and publicly accessible websites and web services. Each of those is a list. An agency could not comply without knowing what was on each list. In our reading, the directive’s first demand was an inventory, even though it never used that word as an action.
Second-level domains
The compliance guide defined a second-level domain as the name an organization registered directly, including the top-level domain. Its example was dhs.gov within www.dhs.gov. Every such domain needed SPF and DMARC records, and later a reject policy. That included domains that never send mail. A parked domain with no DMARC policy can still be spoofed.
An agency may hold domains for programs, campaigns, events and legacy systems. A domain registered for a past initiative and forgotten would still count. The guide noted that OMB memorandum M-17-06 required federal agencies to use the .gov or .mil top-level domains. It told agencies to include any second-level domains not yet on .gov when they reviewed which domains could be preloaded.
Systems operated on an agency’s behalf
The guide said the directive applied to internet-facing agency information systems. That included systems directly managed by the agency and those operated on its behalf. Its focus was agency mail and web infrastructure, regardless of domain suffix. In our reading, a website run by a contractor, a hosted application and a marketing platform sending mail as the agency were all in scope.
That scope matters because those systems can sit outside the CIO’s direct line of sight. A program office might buy a survey tool or newsletter service. A grants program might run a portal through a contractor. Each could send mail or serve web pages under the agency’s name. The directive made the agency answer for each one.
Mail-sending hosts
The reject requirement applied to all second-level domains and mail-sending hosts. We read a mail-sending host as any system that sends mail with the agency’s domain in the From address. In our reading, the hardest part of the inventory was here. Agencies would discover many senders by looking at DMARC reports, which list the sources that reporting receivers saw sending mail that claimed the domain.
Inventory as a living record
Domains are registered and retired. Senders are added when an office buys a new tool. Websites launch and are decommissioned. A one-time inventory would decay. In our reading, the directive pushed agencies toward a maintained register of domains, senders and sites, with an owner for each. That register would serve every later control that depends on knowing what an agency exposes.
Domains that send no mail
A domain that sends no mail is still a target. A spoofer can put it in a From address, and receivers have no policy telling them to refuse the message. Protecting such a domain is simple in principle. Its owner publishes an SPF record that authorizes no senders and a DMARC policy at reject. In our reading, the directive’s rule that every second-level domain needed DMARC covered these dormant assets, which an inventory built from mail traffic would miss.
The same applies to subdomains. DMARC lets a domain owner set a policy for subdomains as well as the domain itself. A spoofer could otherwise invent a subdomain that never existed. The directive required reject for all second-level domains and mail-sending hosts. In our reading, an agency would need to confirm that its policy also covered subdomains it did not use.
Registration and retirement
The inventory is easiest to keep at the two moments a domain changes state. When an office asks for a new domain, the request can carry the owner, purpose, mail plan and web plan. When a domain is retired, the owner can decide whether to keep it parked with protective records or let it go. In our reading, a domain that lapses and is registered by someone else is a larger risk than one kept parked, since links and old mail may still point to it.
Delegated zones
Some agencies delegate subdomains to components, programs or contractors, who run their own DNS for them. Records in a delegated zone are set by whoever holds the delegation. An agency that did not track delegations could not be sure what policies its subdomains published. In our reading, delegation records belong in the inventory, with the name of the party that controls each zone.
Who owns the inventory
The inventory crosses offices. Domain registration may sit with a web team or communications office. Mail systems sit with infrastructure. Web applications sit with program offices and their contractors. Procurement holds the contracts that bring new senders. In our reading, a CIO would need each of these offices to report changes, and a single owner to keep the register current.
The mail path: from STARTTLS and p=none to a reject policy in one year
The email requirements came in stages. The first stage asked for things that would not disrupt mail. The last asked for a setting that could block legitimate mail if the inventory was wrong. The staging gave agencies time to learn before enforcement.
STARTTLS on receiving servers
The directive required all internet-facing mail servers to offer STARTTLS within 90 days. The background section explained that STARTTLS does not force encryption. It signals that encryption is available, which makes passive interception harder. In our reading, this was the least disruptive step, since a sender that cannot use encryption will still deliver.
SPF and DMARC at p=none
Within the same 90 days, every second-level domain needed valid SPF and DMARC records, with a DMARC policy of at least p=none and a reporting address. A p=none policy tells receivers to take no special action on failing mail but to send reports. It is a monitoring mode. The directive used it to give agencies data before enforcement.
The reports are the value of p=none. Receivers that support DMARC send aggregate reports listing the sources that sent mail claiming the domain and whether each passed. An agency reading those reports would see its own mail servers, its vendors and any spoofers. The directive’s staging assumed agencies would use the months between p=none and reject to work through that list.
From p=none to reject
Within one year, agencies had to set p=reject on all second-level domains and mail-sending hosts. Under reject, receivers that honor DMARC refuse mail that fails authentication. A legitimate sender that had not been brought into SPF or DKIM alignment would have its mail rejected. The directive’s background said reject provides the strongest protection against spoofed email.
The work between the two stages was alignment. DMARC requires that the domain in the visible From address align with a domain authenticated by SPF or DKIM. The guide explained that alignment can be strict or relaxed. For each vendor sending as the agency, the agency would need to publish the vendor’s sending sources in SPF, or have the vendor sign mail with a DKIM key for the agency’s domain.
Third-party senders
Vendors are where the work concentrated, in our reading. A notification service, a payroll provider, a benefits portal or a survey platform might each send mail as the agency. Each would need configuration on the vendor side and DNS changes on the agency side. Some vendors might not support DKIM signing for a customer domain. In those cases, the agency would need to change the From address to a vendor domain or a subdomain, or change vendors.
That is a procurement question as well as a technical one. Contracts signed before the directive would not mention DMARC. New contracts could require vendors to support SPF and DKIM alignment for agency domains. In our reading, the directive’s budget note to CIOs and procurement officers pointed at exactly this kind of cost.
Reading the reports takes tooling
Aggregate reports arrive as machine-readable files from each receiving organization that sends them. A busy domain can receive reports from a long list of receivers every day. Reading them by hand does not scale. In our reading, an agency would need a tool or service to collect, parse and summarize the reports by sending source. Choosing that tool was another procurement step the directive implied without naming. The tool would also hold data about the agency’s mail flows, which deserves the same care as other security logs.
Forwarding and mailing lists
DMARC has known friction points. A message forwarded by an intermediary may fail SPF, because the forwarder’s server is not in the sender’s SPF record. A mailing list that changes the subject or body may break the DKIM signature. Under a reject policy, such mail can be refused. In our reading, agencies with staff on external mailing lists, or with mail forwarded through partner systems, would see these failures in their reports and need to plan for them.
Not every failure is a problem to fix on the agency’s side. Some sit with the intermediary. The aggregate reports show which source failed and why. That lets an agency tell the difference between its own misconfigured vendor and a third party’s forwarding. The directive’s one-year staging gave time to sort the two.
Inbound mail was a separate question
The directive concerned mail that claims agency domains. It did not require agencies to enforce other organizations’ DMARC policies on mail they received. Whether an agency’s own gateway rejected spoofed mail claiming a bank or a vendor was outside its required actions. In our reading, the directive protected the public and other receivers from mail claiming to be the government. Protecting agency staff from mail claiming to be someone else needed inbound controls the directive did not set.
DKIM keys as managed assets
DKIM relies on a private key that signs mail and a public key published in DNS. Each sending service may use its own key. Keys need generation, publication, rotation and retirement. A vendor that leaves a contract may leave a public key in the agency’s DNS. In our reading, the inventory of senders should list their DKIM selectors, so an agency can remove keys when a service ends.
Reports to DHS
The directive required agencies to add DHS as a recipient of DMARC aggregate reports once the central reporting location was set up. The compliance guide gave the address as reports@dmarc.cyber.dhs.gov. In our reading, this gave DHS a cross-government view of mail claiming federal domains. It could see spoofing campaigns against several agencies at once, and see which agencies still had unaligned senders.
Legacy ciphers on mail servers
The directive also required disabling SSLv2, SSLv3, 3DES and RC4 on mail servers. Mail flows between organizations with different software ages. Turning off old ciphers could break delivery with partners running old servers. The directive’s page later noted a temporary policy exception for 3DES in mail environments, which we return to in hindsight.
The web path: HTTPS everywhere, HSTS, preloading and retired ciphers
The web requirements built on M-15-13. That memorandum had already required HTTPS-only service with HSTS for existing federal websites and web services. The guide said the directive incorporated parallel language. It added two steps: disabling old protocols and ciphers, and reviewing which domains could be preloaded.
HTTPS-only with HSTS
Within 120 days, all publicly accessible federal websites and web services had to serve only over HTTPS, with HSTS. The background said HTTP connections can be monitored, modified and impersonated, and that HSTS makes browsers always use HTTPS and removes the ability to click through certificate warnings. In our reading, that last effect matters as much as encryption. A user cannot override a misconfigured certificate on an HSTS site.
Web services include application programming interfaces. An agency API used by states, partners or the public would need HTTPS-only service too. API clients written against plain HTTP endpoints would break when the endpoint stopped answering on HTTP. In our reading, agencies would need to tell their API users before the switch, and some users would be other governments.
Preloading
The directive asked agencies to identify second-level domains that could be HSTS preloaded and send DHS the list. The guide explained that preloading enforces HTTPS across an entire zone and counts as technical compliance with the HTTPS requirements. It lets an agency avoid configuring HSTS for every subdomain. It also affects every subdomain, including internal ones.
That is where preloading becomes an architecture question. A domain used for both public sites and internal intranet hosts cannot be preloaded until every internal host supports HTTPS. Preloading would make browsers refuse plain HTTP to any host under the domain. In our reading, the directive’s preload list was also a map of where internal systems still used plain HTTP.
Retired protocols and ciphers
The guide gave the history. SSLv2 dates from 1995, and the DROWN attack showed that merely serving it exposes traffic encrypted with newer TLS versions. SSLv3, from 1996, was considered insecure after the POODLE attack was published in 2014. In 2014, NIST marked RC4 as not approved for federal systems. In 2017, NIST urged users of 3DES to migrate as soon as possible. The directive required that these cease to be offered on internet-facing web and mail servers. It noted that turning off SSLv3 removes support for Internet Explorer 6.
Disabling a cipher is a configuration change. Knowing what breaks is not. Agencies with old clients, appliances or partner systems would need to test. In our reading, the cipher requirement exposed old dependencies the same way the reject requirement exposed unknown senders.
Certificates become a dependency
HSTS changes the cost of certificate errors. Without HSTS, a user facing an expired certificate can click through a warning. With HSTS, the browser refuses the connection. A lapsed certificate becomes an outage. In our reading, the directive made certificate inventory and renewal an availability control. An agency would want automated renewal where possible, and an alert well before expiry where not.
Preloading extends this to every host under a domain. A forgotten internal server with an expired or self-signed certificate would become unreachable in browsers once its parent domain was preloaded. The risk is one reason preloading moves slowly. It also explains why the preload list doubles as a list of hosts that need certificate management.
Non-browser clients and old links
HSTS protects browsers. It does not change how scripts, devices and partner systems connect. A data feed fetched over plain HTTP by another agency’s script would simply fail when the server stopped answering on HTTP. Old links in printed forms or archived documents may point to HTTP addresses. In our reading, agencies would need redirects for browsers, notice for machine clients, and a period of logging to find who still called HTTP endpoints.
Load balancers, CDNs and hosted platforms
Web traffic can terminate at a load balancer, content delivery network or hosted platform rather than at the agency’s server. The protocols and ciphers offered to the public are set there. For a site on a hosted platform, the agency might not control those settings directly. In our reading, the directive made those settings a contract matter, since the agency answered for them.
The evidence: external scans, DMARC reports and monthly status reports
The directive used three kinds of evidence. DHS scanned agencies’ internet-facing systems. DHS received DMARC aggregate reports. Agencies sent plans and status reports every 30 days until done. The first two did not depend on what the agency said.
Scanning as verification
The directive said DHS would rely on scanning to track and verify progress. Every required setting was externally observable. A scan could read a domain’s DMARC record, connect to its mail server and see whether STARTTLS was offered, and connect to its website and list protocols, ciphers and headers. In our reading, an agency’s compliance status would be whatever the scan saw.
That has consequences for how an agency manages compliance. If DHS scans a domain the agency forgot, the forgotten domain fails. If a contractor changes a load balancer setting, the agency fails the next scan. Compliance is continuous, not a point-in-time attestation. An agency would want to run the same scans on itself, more often than DHS, to see failures first.
DMARC reports as a second view
DMARC reports gave DHS a view no scan could. They show who is actually sending mail claiming each domain. An agency that set p=reject without fixing a vendor would see rejections in its reports. DHS would see them too. In our reading, the reports made the mail path observable in operation as well as in configuration.
Status reports and plans
The plan of action and the status reports were the traditional element. They let agencies explain constraints and timelines. But a plan claiming completion would not survive a scan showing otherwise. In our reading, the written reports became context for the scan results, not a substitute for them.
What external evidence cannot show
External scans see only what faces the internet. They cannot see whether internal web traffic is encrypted, whether internal mail relays authenticate, or whether staff can still reach internal sites over HTTP. The directive did not reach those. In our reading, its evidence model fit its scope. Internal traffic would need different evidence, and later policy would ask for it.
Public visibility
The same scans that DHS ran could be run by anyone. A researcher, journalist or adversary could check an agency’s DMARC policy in seconds. In our reading, that made compliance visible beyond the government. An agency lagging on reject would be visibly lagging. That visibility added pressure the directive did not need to write down.
Metrics an agency could keep
The directive’s settings translate into a small set of measures. The share of second-level domains at DMARC reject. The number of sending sources in aggregate reports that fail alignment. The share of public hosts serving HTTPS only with HSTS. The number of hosts still offering retired protocols or ciphers. The number of domains on the preload list. Each can be computed from external observation and the agency’s register.
Measures like these let a CIO report progress in the same terms DHS used. They also reveal drift. A domain that falls back to p=none, or a new host without HSTS, shows up as a change in a number. In our reading, an agency that tracked these measures weekly would find regressions before the next external scan.
Exceptions need records
Not every system could comply on time. A legacy application might need an old cipher for a partner. A vendor might not support DKIM signing. The directive’s 30-day status reports were the channel to explain these. In our reading, an agency would want a written exception for each case, with a reason, an owner, compensating controls and a date to revisit. Without that record, an exception becomes permanent by neglect.
Inspectors general and auditors
A directive with measurable settings is easy to audit. An inspector general can scan the same systems DHS scans and compare results with an agency’s reports. In our reading, the directive gave auditors a clear test of whether an agency’s reporting was accurate. That raised the cost of optimistic status reports.
How the directive reached different organizations
The directive bound federal civilian executive branch agencies. Its effects reached further, through vendors, partners and the example it set.
Cabinet departments
A large department may hold domains across several components. Each component may run its own mail and web infrastructure. In our reading, a department CIO would need component inventories rolled up, a common approach to vendor senders, and a way to show DHS progress across the department. The 30-day status cycle would test that roll-up each month.
Small agencies, boards and commissions
Small agencies were bound by the same deadlines with smaller staffs. A small agency may rely on a parent department or a shared service provider for mail and web hosting. In our reading, it would then depend on that provider to make changes, and would need the provider’s commitment in writing. Their inventories would be shorter, but each item would weigh more against a small team.
Contractors and software vendors
Vendors sending mail or hosting sites for agencies would face requests to support SPF, DKIM alignment, HTTPS-only service and modern ciphers. The directive did not bind them directly. Their federal customers would ask, and contracts would follow. In our reading, vendors that sold to several agencies would find it cheaper to support these settings once for all customers.
Shared service providers
Providers that host mail or websites for several agencies would carry the directive’s requirements for all of them at once. A change to a shared mail gateway’s cipher list affects every tenant. In our reading, a shared provider would need a single plan covering its tenants, and each tenant would need evidence of that plan for its own reporting. The directive’s reporting cycle would expose any gap between provider and tenant.
Contractors’ own domains
Contractors communicate with agencies from their own domains. The directive did not reach those. An agency’s staff could still receive spoofed mail claiming to come from a contractor. In our reading, agencies that wanted protection on that path would need to ask contractors to publish DMARC on their domains, through contract terms or supplier security requirements. The directive set an example contractors could follow without being bound.
State and local governments
The directive did not apply to state and local governments. They also hold .gov domains, and their residents receive mail from them. In our reading, a state or county that saw the directive would recognize the same risks to its own domains. The .gov registry’s later guidance, discussed in hindsight, recommended DMARC to all .gov holders.
The public as recipients
Residents receive mail from agencies about benefits, taxes, licenses and appointments. A message that claims an agency’s exact domain and fails authentication would, under a reject policy, be refused by receivers that honor DMARC. In our reading, that protects people who may act on an official-looking message. It does not protect them from lookalike domains or from messages that use display names alone. Agencies would still need to tell the public how to recognize real messages, and which domains they use.
HTTPS-only service protects the same people on the web. A resident filling in a form on an agency site would reach it over an encrypted connection, with no certificate warning to click through. The directive’s web actions reached the public directly, even though it addressed only agencies.
Banks and health systems
Banks and health systems receive mail from federal agencies and send mail to the public under their own domains. When federal domains moved to p=reject, mail spoofing an agency would be refused by receivers that honor DMARC. In our reading, private institutions could take two lessons. Their own domains face the same spoofing risk. And the staged approach, monitoring first and enforcement later, applies to them as well.
Objections, in order
The argument that the directive made internet-facing configuration the compliance record, and that this approach outlasted its settings, faces objections. We take the strongest six in order.
1. These were basic settings, so the directive added little
The settings were well known. The directive’s figure was that seven of the ten most common vulnerabilities in federal networks would be addressed by its web actions. If the settings were basic, the DHS figure suggests they were also missing across agencies. The directive’s contribution was to set dates and a verification method for basics that had not been done.
2. External scans miss the real risk inside networks
They do miss internal risk. The directive did not claim otherwise. Its scope was internet-facing mail and web systems, where scanning works well. The objection supports our point that the evidence model fits the scope, and that internal controls need a different method.
3. DMARC reject breaks legitimate mail
It can, if senders are not aligned. That is why the directive staged p=none first, with reports. An agency that used the year to align senders would reject spoofed mail and keep legitimate mail. The risk falls on agencies that skipped the monitoring stage, not on the policy.
4. Spoofers can use lookalike domains instead
They can. DMARC protects the exact domain, not similar ones. The directive did not address lookalike domains. It closed the easiest path, mail that claims the real domain. Lookalike domains need different controls, such as monitoring registrations and educating users.
5. Compliance by scan rewards checklists, not security
Any measurable standard can be met minimally. But the directive’s measures were the controls themselves, not proxies for them. A domain with p=reject rejects spoofed mail. A site with HSTS refuses plain HTTP. In our reading, when the measured setting is the protection, meeting the measure provides the protection.
6. Cloud email made the directive obsolete
Agency mail can run on cloud platforms, which have their own security features. But DMARC, SPF and DKIM are domain settings, not server features. They apply wherever mail is hosted. CISA’s later cloud baseline, discussed in hindsight, kept the same DMARC requirements for cloud mail tenants. The platform changed. The domain-level evidence stayed.
What an agency CIO or CISO could check
The directive’s specific deadlines are long past. Its approach, observable configuration as evidence, applies to any organization with domains and web services. The history suggests a short set of checks:
- Keep a register of every second-level domain, mail-sending service and public website, with a named owner for each.
- Read DMARC aggregate reports regularly and reconcile each sending source against the register.
- Confirm that every domain, including parked ones, has a DMARC policy at reject and an SPF record.
- Require vendors that send mail or host sites for the organization to support DKIM alignment, HTTPS-only service and current ciphers.
- Scan your own internet-facing systems more often than any outside party does, and fix failures before they are reported.
- List domains that cannot be preloaded and the internal systems that block them.
The checks apply beyond federal agencies. A bank, health system or state agency has the same exposure: domains that can be spoofed, vendors that send in its name, and websites whose protocols anyone can read. The same staged approach works for each. Start with an inventory, publish monitoring policies, read the reports, align or replace senders, then enforce. The order matters more than the speed, because enforcement before inventory blocks legitimate mail.
These checks give a leader numbers to report: domains at reject, unaligned senders, sites without HSTS, domains not preloaded. Each can be verified by anyone with a scanner. In our view, that is the standard an organization’s public-facing security should meet: evidence that does not depend on the organization’s own word.
With hindsight
The directive’s page on CISA’s site records one adjustment to its settings. It notes that DHS acknowledged a significant constraint around disabling 3DES in mail environments. It issued a temporary policy exception on that requirement on September 20, 2018.
Federal policy then moved inward. On January 26, 2022, OMB issued memorandum M-22-09, Moving the U.S. Government Toward Zero Trust Cybersecurity Principles. It set goals for agencies to meet by the end of fiscal year 2024. It required agencies to enforce HTTPS for all web and API traffic in their environments, including traffic that is solely internal. It required agencies to work with CISA to preload their .gov domains as HTTPS-only in browsers.
The memorandum described the state of preloading. It said that since 2020 the DotGov program had coordinated with browsers to preload all newly registered .gov domains. It said many existing agency .gov domains had not been preloaded. The most significant barrier, it said, was intranet websites on publicly registered .gov domains that did not support HTTPS. It noted the .gov registry’s stated intent to preload the entire .gov space eventually.
M-22-09 also turned to mail. It said CISA would work with FedRAMP to evaluate government-wide solutions for encrypted email in transit and make recommendations to OMB. In our reading, that conceded a gap the 2017 directive had left. STARTTLS signals that encryption is available but, as the directive itself said, does not force it.
The .gov registry’s site, get.gov, is an official CISA website. As checked on November 15, 2026, its domain security page says all newly registered .gov domains are preloaded. It says the registry intends to preload the .gov top-level domain. It recommends that .gov holders preload existing domains, noting this is a required action for federal agencies under the zero trust strategy. It also recommends a strong DMARC policy, even for domains that do not send email.
On December 17, 2024, CISA issued BOD 25-01, Implementing Secure Practices for Cloud Services. It required agencies to inventory in-scope cloud tenants, deploy CISA’s assessment tools and report results continuously or quarterly. It required mandatory configuration policies by June 20, 2025. Authorizing officials could accept risk for deviations, which agencies had to explain in their reports.
CISA’s list of required configurations under BOD 25-01 carries BOD 18-01’s mail settings forward. For Exchange Online, a DMARC policy shall be published for every second-level domain. The policy shall be p=reject. The aggregate report contacts shall include reports@dmarc.cyber.dhs.gov. For Google Workspace, the list requires a DMARC policy at the full domain or the second-level domain for all Workspace domains. In our reading, the 2017 directive’s mail requirements became part of a cloud configuration baseline checked by automated tools.
As checked on November 15, 2026, CISA’s page for BOD 18-01 shows no revocation notice, while the same page’s list of related directives marks BOD 19-02 and BOD 22-01 as revoked. What held up is the evidence model. The settings were extended inward by M-22-09 and carried into BOD 25-01. In both, compliance is shown by configuration that tools can read, not by an agency’s statement.
Spotted an error? Tell us and we will check it and correct the article where needed.