defaultsexposed/domain-security-census

Domain Security Census

Monthly aggregates over the registered-domain population: 332.5M graded domains across 1,434 TLDs, as of 12 October 2026.

Homepage: https://defaults.exposed/data. Source of record: Zenodo, https://doi.org/10.5281/zenodo.22690388. Methodology: https://defaults.exposed/methodology.

Edition 2026-10, as of 12 October 2026. Licence CC BY 4.0. Concept DOI 10.5281/zenodo.22690388.

The Domain Security Census grades the registered-domain population on the public signals that decide whether a domain's email can be forged and whether its website is safe to visit. This dataset holds edition 2026-10 as 18 CSV tables (24,184 rows), flattened from the 12 aggregate JSON files deposited on Zenodo. Every figure is a count or a share of counts. No file names a domain, and no row describes a single domain.

Coverage: a census of the registered-domain population, not a sample

This edition takes every registered domain name its three sources list, scans each one, and grades each one that exists. As of 12 October 2026:

Stage	Count
Names sourced from ICANN CZDS zone files, OpenINTEL and CommonCrawl, deduplicated	649,194,272
Names scanned	632,901,201
Names that exist (NXDOMAIN answers removed)	361,301,790
Domains graded	332,492,319
TLDs scanned	1,434
TLDs broken out one by one in the per-TLD tables	520

Covered. Every name left in the merged feeds once wildcard phantoms and duplicate encodings are removed, across 1,434 TLDs, scanned from 11 October 2026 to 12 October 2026. Every graded domain counts in every total, including the 711,046 graded domains in the 914 TLDs that are not broken out on their own.

Outside the census.

• Registered names that none of the three feeds lists.
• Names that answered NXDOMAIN (271,599,411): listed in a feed and no longer registered. They are reported in the basis below and never graded.
• Wildcard phantoms (16,500,126): names a wildcarded TLD answers for whether or not anyone registered them. The 9 wildcarded TLDs are excluded wholesale, and 235,507 names under them are recovered from the feeds and scanned.
• Anything about an individual domain. The files hold counts and shares only.

Scope. Every figure is a full count over the scanned population, so the figures carry no sampling error. The count covers every name the three sources list. Registered names that none of them lists are outside the census. 28,809,471 names that exist were scanned and reported as indeterminate (15,201,942) or unreachable (13,607,529); they are counted separately from the graded base, in the basis below.

Why population coverage matters for reuse. The per-TLD tables come from the same scan, so a small ccTLD is measured with the same completeness as .com and the two can be set side by side. Rare states are counted in full: a misspelt DMARC policy or an SPF record ending +all appears with its exact count.

Citation

Cite the concept DOI. It resolves to the newest edition and covers every edition.

Defaults Exposed FZ-LLC. (2026). Domain Security Census: Aggregate Dataset (Version 2026-10) [Data set]. Zenodo. https://doi.org/10.5281/zenodo.22690388

@dataset{defaults_exposed_domain_security_census,
  author    = {{Defaults Exposed FZ-LLC}},
  title     = {Domain Security Census: Aggregate Dataset},
  year      = {2026},
  version   = {2026-10},
  publisher = {Zenodo},
  doi       = {10.5281/zenodo.22690388},
  url       = {https://doi.org/10.5281/zenodo.22690388}
}

To fix this edition's figures, add its version DOI, 10.5281/zenodo.23368770. Licence: Creative Commons Attribution 4.0 International (CC BY 4.0), https://creativecommons.org/licenses/by/4.0/. Reuse for any purpose, commercial work included, with attribution.

Where this dataset lives

The same edition is published in three places from one write-up.

• Zenodo, the source of record: the census's aggregate JSON files, this README, a JSON data dictionary and CITATION.cff. https://doi.org/10.5281/zenodo.22690388
• Kaggle: the same figures as 18 CSV tables, with a starter notebook. https://www.kaggle.com/datasets/defaultsexposed/domain-security-census
• Hugging Face: the same 18 CSV tables, each browsable in the dataset viewer. https://huggingface.co/datasets/defaultsexposed/domain-security-census
• defaults.exposed: current figures, downloads and the narrated statistics. https://defaults.exposed/data
• Methodology: every check, its pass and fail conditions and its weight. https://defaults.exposed/methodology

What the census is

The census grades each domain on SPF, DMARC, DKIM, MX, reverse DNS, DNSSEC, CAA, name servers, SOA, HTTPS, TLS, certificates, IPv6 and the main HTTP security headers. Each domain is graded A+ to F on the published methodology at https://defaults.exposed/methodology. A check that cannot be determined is left out of that domain's score.

Names come from three feeds: ICANN CZDS zone files, OpenINTEL, and CommonCrawl (crawled hosts and the webgraph). The feeds are merged and deduplicated before scanning. The scan ran from 11 October 2026 to 12 October 2026 and produced one result row per domain. The published files aggregate those rows by check, by state and by top-level domain (TLD).

The basis

From sourced names to graded domains

Stage	Names	What happens
Names sourced	649,194,272	Deduplicated across CZDS, OpenINTEL and CommonCrawl
Wildcard phantoms removed	16,500,126	Names a wildcarded TLD answers for whether or not they are registered (9 wildcarded TLDs excluded wholesale)
Duplicate IDNA encodings removed	28,452	Two spellings of one internationalised name collapsed to one
Wildcard-TLD names added	235,507	Names under wildcarded TLDs recovered and carried into the scan
Names scanned	632,901,201	Every name the scan covered
Answered NXDOMAIN	271,599,411	In the sources and no longer registered. Reported separately and never graded
Names that exist	361,301,790	The scanned basis of this edition (lifecycle.json total)
Disposition graded	329,604,717	The checks ran and a grade was computed
Disposition indeterminate	15,201,942	A DNS query the verdict depends on failed (SERVFAIL, timeout), so no grade is assigned
Disposition unreachable	13,607,529	The name exists and the scan could not reach it
Disposition dead	2,887,602	The name exists, its NS query returned a definite empty answer, and no address or mail record resolves
Graded	332,492,319	Disposition graded plus the existing-but-dead names, which grade F. The headline basis

The headline is drawn from names that exist. A name that answers NXDOMAIN is no longer a domain and is not graded. A name that exists and is dead is graded F.

Every denominator

Each share is a count divided by one of these bases. Divide by the base the table or file names, and name that base when you quote the share.

Base	Count	What it counts
Names that exist	361,301,790	Scanned names that exist
Graded	332,492,319	Every grade share; the n of dmarc, spf, mx, dnssec, https and ipv6; every overlap count; MX presence
Disposition graded	329,604,717	The checks ran; the n of ns_diversity (dead names excluded)
HTTPS-serving	220,264,382	The n of tls, certificate and hsts; TLS versions; certificates observed
Header responders	132,559,479	The n of csp, clickjacking, mime_sniffing and referrer_policy: domains that returned a 2xx page the header checks evaluated
All five header checks evaluated	115,867,940	The n of secure_by_default
DKIM evaluated	76,563,365	The n of dkim: MX present and the provider recognised, so its selectors are known
MX with a resolvable primary host	180,346,153	The n of reverse_dns
CAA answered	328,504,222	The n of caa: the lookup returned a determinate answer
Certificate checks	220,258,401 / 220,271,759 / 220,271,422	The n of cert_expiry, cert_sigalg and cert_key, each its own applicable population
HTTP redirect check	211,262,083	The n of https_redirect
SOA check	327,527,446	The n of soa
DMARC record holders	83,466,694	DMARC policy shares
SPF record holders	157,065,571	SPF ending shares
Domains with MX (raw MX list)	180,564,746	Mail provider shares
Domains with NS	326,448,798	DNS operator shares
Server header present	220,432,333	Web server shares

Edition highlights

Every figure below is as of 12 October 2026.

Grades

• 247,066,987 of 332,492,319 graded domains grade F (74.3%). 87,562 reach A+ (0.03%).
• By kind, over the 520 broken-out TLDs: generic TLDs grade 77.5% F (234,547,634 graded across 353 gTLDs); country-code TLDs grade 66.7% F (97,233,639 graded across 167 ccTLDs).
• .com grades 73.9% F (153,849,480 graded) and .nl 41.4% F (4,055,303 graded).

Email

• 39,758,587 graded domains enforce DMARC with p=quarantine or p=reject (12.0%); 18,601,119 of them use p=reject (5.6%).
• 91,705,195 graded domains publish SPF and no DMARC record (27.6%).
• 43,658,833 of 83,466,694 DMARC records are p=none (52.3%), and 32,980,787 carry a rua= reporting address (39.5%). 44,239 records are malformed: 27,170 have no p= tag, and the most common misspelling of quarantine, quarentine, appears in 3,840.
• 87,173,390 of 157,065,571 SPF records end ~all (55.5%) and 61,408,709 end -all (39.1%). 40,988 end +all, which authorises every server on the internet to send as the domain. 8,317 need more than 10 DNS lookups.
• 180,564,746 of 332,492,319 graded domains publish MX (54.31%). Google Workspace receives mail for 22,069,386 of them (12.22%) and Microsoft 365 for 14,967,920 (8.29%). 96,688,981 (53.55%) point their primary MX at a host the provider map does not name.

Web and TLS

• 208,230,719 of 220,264,382 HTTPS-serving domains negotiate TLS 1.3 (94.5%) and 12,033,663 negotiate TLS 1.2 (5.5%). By TLD: .uk 96.2% (5,570,039 HTTPS-serving), .com 94.9% (101,602,779), .nl 91.5% (3,356,283).
• 47,116,222 of 220,264,382 HTTPS-serving domains send HSTS (21.4%).
• 196,381,007 certificates validate (89.2%) and 23,883,375 fail (10.8%). 4,390,979 are self-signed: 2.0% of all certificates observed and 18.4% of the failures. 9,355,930 are expiring or expired (4.2% of the cert_expiry n).
• Let's Encrypt issued 124,837,486 of the 220,264,382 certificates observed (56.68%), Google Trust Services 39,207,201 (17.80%) and GoDaddy 26,754,374 (12.15%).
• 220,432,333 graded domains send a Server header (66.3%). cloudflare leads with 21.33% of them, then nginx (17.61%) and apache (16.10%). 25,795,041 send X-Powered-By (7.8% of graded); 14,541,793 of those name a PHP version (56.4%).

DNS and exposure

• 23,390,098 graded domains are DNSSEC signed and valid (7.0%).
• GoDaddy runs DNS for 53,313,849 domains (16.33%) and Cloudflare for 52,173,900 (15.98%). The five largest operators together hold 42.8%.
• 60,347,767 graded domains hold none of five controls (SPF, enforcing DMARC, DNSSEC, HTTPS, HSTS), 18.2%; 1,325,934 hold all five (0.40%).

Lifecycle

• 329,604,717 of 361,301,790 names that exist were graded (91.2%). 15,201,942 were indeterminate (4.2%) and 13,607,529 unreachable (3.8%).
• 2,887,602 exist and are dead (0.80%). They grade F and count in the 332,492,319 graded base.

Reading the numbers

• Each edition is a snapshot of its own scanned population. Sources, the TLD mix and the method move between editions, so a difference between two editions' shares measures those changes along with any change in the internet. Shares are not differenced as a trend. The method notes below set out what moved this edition.
• Each check has its own denominator, and the table above lists every base. TLS, HSTS and certificate figures are shares of HTTPS-serving domains; header figures are shares of header responders.
• .com holds 153,849,480 of the 332,492,319 graded domains, so all-TLD figures lean towards it. Compare TLDs on the generic and country-code split.
• DMARC is parsed twice. The grading engine's parse (checks.json dmarc, the one each grade is computed from) and a re-parse of the published records (dmarc-policies.json) both find 83,466,694 records. Their policy splits differ: p=none is 43,680,223 in the first and 43,658,833 in the second. Quote each with its file.
• The MX check counts 180,462,439 domains with MX; the raw MX list behind the provider shares counts 180,564,746. Quote each with its file.
• ca-share.json, cert-problems.json, ns-providers.json and stack-headers.json count scan rows, 1.1% of which are re-scan duplicates. Shares are unaffected.
• 8 Server values ending .compute.internal are the backend hostnames of one domain-forwarding service, which returns each backend's internal name as its Server header. Together they account for 7,491,643 domains; read them as one service.
• The TLS figures report the version negotiated with the scanner. A server that offers TLS 1.0 and agrees TLS 1.3 counts as TLS 1.3. Legacy protocol support is measured in no field.
• https.redirects, tls.tls11, tls.tls10, tls.other, tls.weak_cipher and csp.weak are zero by construction in checks.json, and the CSV tables leave them out. Redirects are measured in https_redirect.
• The IPv6 figure is AAAA presence from the raw DNS answer, scaled from the raw pass to the graded base.
• Every two-letter TLD counts as country-code, including those sold as generic, such as .io and .ai. The CSV tables take kind from the IANA root zone database. tlds.json kind follows IANA for 519 of its 520 entries; xn--j1amh departs from it.

Tables

One CSV per table, UTF-8, comma-separated, header row first. File names stay the same in every edition. Shares are recomputed from the counts in the same row.

File	One row per	Built from	Rows	What it holds
grades_by_tld.csv	TLD	tlds.json	521	A+ to F grade counts for every graded domain (ALL) and for each of the 520 TLDs broken out.
grades_gtld_vs_cctld.csv	TLD kind	tlds.json	2	The same grade counts summed over the broken-out TLDs by kind: generic against country-code.
checks_global.csv	check and state	checks.json	55	Every state of every check across all graded domains, with each check's own denominator.
checks_by_tld.csv	TLD, check and state	checks.json	22,360	Every state of the 17 per-TLD checks for each broken-out TLD.
check_denominators.csv	check	checks.json	23	What each check's denominator counts, with the producer's caveat for the check.
email_web_overlaps_by_tld.csv	TLD	overlaps.json	521	Controls domains hold together, for example SPF without DMARC and HTTPS without HSTS, overall and per TLD.
exposure_score.csv	number of controls	overlaps.json	6	Graded domains by how many of five controls they hold: SPF, enforcing DMARC, DNSSEC, HTTPS and HSTS.
tls_version_by_tld.csv	TLD	tls-versions.json	521	TLS version negotiated with the scanner, overall and per TLD.
dmarc_policy.csv	policy	dmarc-policies.json	4	Policy of every published DMARC record, re-parsed from the record.
dmarc_malformed_policy_tokens.csv	token	dmarc-policies.json	20	The 20 most common p= values in malformed DMARC records.
spf_all_qualifier.csv	qualifier	spf-errors.json	5	How published SPF records end.
mx_providers.csv	provider	mx-providers.json	30	The 30 largest mail providers by primary MX host.
dns_providers.csv	provider	ns-providers.json	30	The 30 largest DNS operators by first NS host.
certificate_issuers.csv	issuer	ca-share.json	26	The 25 largest certificate issuers, plus Other for the rest.
certificate_validity.csv	measure	cert-problems.json	3	Valid, invalid and self-signed certificates.
web_server_header.csv	server	stack-headers.json	25	The 25 most common Server header values. Values ending .compute.internal are the backend hostnames of one domain-forwarding service; read them as one service.
x_powered_by_header.csv	value	stack-headers.json	20	The 20 most common X-Powered-By values.
population.csv	stage	figures sheet and lifecycle.json	12	Every stage from names sourced to domains graded, for this edition.

Data dictionary

grades_by_tld.csv

A+ to F grade counts for every graded domain (ALL) and for each of the 520 TLDs broken out.

Column	Type	Description
tld	string	Top-level domain without the dot, internationalised TLDs in punycode. ALL is every graded domain; ALL_GTLD and ALL_CCTLD are the rollups by kind.
tld_kind	string	gtld (generic) or cctld (country-code), following the IANA root zone database; all for the ALL row.
names_that_exist	integer	Scanned names that exist (NXDOMAIN names removed).
graded	integer	Graded domains: the graded disposition plus names that exist and are dead, graded F.
alive	integer	Names that exist minus dead names.
unreachable	integer	Names that exist that the scan could not reach. Not graded.
grade_A_plus	integer	Domains graded A+.
grade_A	integer	Domains graded A.
grade_B	integer	Domains graded B.
grade_C	integer	Domains graded C.
grade_D	integer	Domains graded D.
grade_F	integer	Domains graded F, including dead names that exist.
f_share_pct	numeric	grade_F / graded x 100, four decimals.
b_or_better_pct	numeric	(grade_A_plus + grade_A + grade_B) / graded x 100, four decimals.

grades_gtld_vs_cctld.csv

The same grade counts summed over the broken-out TLDs by kind: generic against country-code.

Column	Type	Description
tld	string	Top-level domain without the dot, internationalised TLDs in punycode. ALL is every graded domain; ALL_GTLD and ALL_CCTLD are the rollups by kind.
tld_kind	string	gtld (generic) or cctld (country-code), following the IANA root zone database; all for the ALL row.
names_that_exist	integer	Scanned names that exist (NXDOMAIN names removed).
graded	integer	Graded domains: the graded disposition plus names that exist and are dead, graded F.
alive	integer	Names that exist minus dead names.
unreachable	integer	Names that exist that the scan could not reach. Not graded.
grade_A_plus	integer	Domains graded A+.
grade_A	integer	Domains graded A.
grade_B	integer	Domains graded B.
grade_C	integer	Domains graded C.
grade_D	integer	Domains graded D.
grade_F	integer	Domains graded F, including dead names that exist.
f_share_pct	numeric	grade_F / graded x 100, four decimals.
b_or_better_pct	numeric	(grade_A_plus + grade_A + grade_B) / graded x 100, four decimals.

checks_global.csv

Every state of every check across all graded domains, with each check's own denominator.

Column	Type	Description
tld	string	Top-level domain without the dot, internationalised TLDs in punycode. ALL is every graded domain; ALL_GTLD and ALL_CCTLD are the rollups by kind.
tld_kind	string	gtld (generic) or cctld (country-code), following the IANA root zone database; all for the ALL row.
check	string	Check name. Each check is defined at https://defaults.exposed/methodology.
state	string	Outcome of the check for these domains.
count	integer	Domains in this state.
denominator	integer	Domains the check ran on (its n). Each check has its own; check_denominators.csv says what each one counts.
not_applicable	integer	Graded domains the check does not apply to, outside the denominator.
share_pct	numeric	count / denominator x 100, four decimals.

checks_by_tld.csv

Every state of the 17 per-TLD checks for each broken-out TLD.

Column	Type	Description
tld	string	Top-level domain without the dot, internationalised TLDs in punycode. ALL is every graded domain; ALL_GTLD and ALL_CCTLD are the rollups by kind.
tld_kind	string	gtld (generic) or cctld (country-code), following the IANA root zone database; all for the ALL row.
check	string	Check name. Each check is defined at https://defaults.exposed/methodology.
state	string	Outcome of the check for these domains.
count	integer	Domains in this state.
denominator	integer	Domains the check ran on (its n). Each check has its own; check_denominators.csv says what each one counts.
not_applicable	integer	Graded domains the check does not apply to, outside the denominator.
share_pct	numeric	count / denominator x 100, four decimals.

check_denominators.csv

What each check's denominator counts, with the producer's caveat for the check.

Column	Type	Description
check	string	Check name. Each check is defined at https://defaults.exposed/methodology.
denominator	integer	Domains the check ran on (its n). Each check has its own; check_denominators.csv says what each one counts.
not_applicable	integer	Graded domains the check does not apply to, outside the denominator.
denominator_note	string	What the denominator counts, in plain words.
caveat	string	The producer's caveat for this check, where it has one.

email_web_overlaps_by_tld.csv

Controls domains hold together, for example SPF without DMARC and HTTPS without HSTS, overall and per TLD.

Column	Type	Description
tld	string	Top-level domain without the dot, internationalised TLDs in punycode. ALL is every graded domain; ALL_GTLD and ALL_CCTLD are the rollups by kind.
tld_kind	string	gtld (generic) or cctld (country-code), following the IANA root zone database; all for the ALL row.
graded	integer	Graded domains: the graded disposition plus names that exist and are dead, graded F.
email_spf	integer	Domains publishing SPF.
email_dmarc_any	integer	Domains publishing a DMARC record with any policy.
email_dmarc_enforce	integer	Domains with DMARC p=quarantine or p=reject.
email_spf_but_no_dmarc	integer	Domains with SPF and no DMARC record.
email_spf_but_not_enforcing	integer	Domains with SPF and no enforcing DMARC policy.
email_spf_and_dmarc	integer	Domains with SPF and a DMARC record.
email_neither_spf_nor_dmarc	integer	Domains with neither SPF nor DMARC.
email_mx	integer	Domains publishing MX.
email_mx_but_no_auth	integer	Domains publishing MX with no email authentication record.
email_mx_spoofable	integer	Domains publishing MX without an enforcing DMARC policy.
email_spoofable_no_enforcing_dmarc	integer	Domains without an enforcing DMARC policy.
web_https	integer	Domains serving HTTPS.
web_https_but_no_hsts	integer	HTTPS-serving domains without HSTS.
web_https_but_not_secure_by_default	integer	HTTPS-serving domains that do not pass all five header checks.
web_http_only	integer	Domains with no HTTPS service.
web_web_secure_https_hsts_csp	integer	Domains with HTTPS, HSTS and an effective CSP together.

exposure_score.csv

Graded domains by how many of five controls they hold: SPF, enforcing DMARC, DNSSEC, HTTPS and HSTS.

Column	Type	Description
controls_in_place	string	How many of the five controls the domain holds, from 0_fully_exposed to 5_fully_locked.
domains	integer	Number of domains.
share_pct	numeric	domains / graded x 100, four decimals.

tls_version_by_tld.csv

TLS version negotiated with the scanner, overall and per TLD.

Column	Type	Description
tld	string	Top-level domain without the dot, internationalised TLDs in punycode. ALL is every graded domain; ALL_GTLD and ALL_CCTLD are the rollups by kind.
tld_kind	string	gtld (generic) or cctld (country-code), following the IANA root zone database; all for the ALL row.
tls_reachable	integer	HTTPS-serving domains that completed a TLS handshake.
tls13	integer	Domains that negotiated TLS 1.3.
tls12	integer	Domains that negotiated TLS 1.2.
tls13_share_pct	numeric	tls13 / tls_reachable x 100, four decimals.

dmarc_policy.csv

Policy of every published DMARC record, re-parsed from the record.

Column	Type	Description
policy	string	p= value: none, quarantine, reject, or malformed (a v=DMARC1 record whose p= is none of the three).
domains	integer	Number of domains.
share_of_records_pct	numeric	domains / DMARC records x 100, four decimals.

dmarc_malformed_policy_tokens.csv

The 20 most common p= values in malformed DMARC records.

Column	Type	Description
token	string	The p= value as published. (no-p-token) marks a record with no p= tag.
domains	integer	Number of domains.

spf_all_qualifier.csv

How published SPF records end.

Column	Type	Description
qualifier	string	hardfail_dash_all (-all), softfail_tilde_all (~all), neutral_qmark_all (?all), plus_all (+all) or other_or_none (no such terminal).
domains	integer	Number of domains.
share_of_spf_pct	numeric	domains / domains publishing SPF x 100, four decimals.

mx_providers.csv

The 30 largest mail providers by primary MX host.

Column	Type	Description
provider	string	Operator of the first MX host. Self-hosted / other groups hosts the provider map does not name.
domains	integer	Number of domains.
share_of_mx_domains_pct	numeric	domains / domains publishing MX x 100, two decimals.

dns_providers.csv

The 30 largest DNS operators by first NS host.

Column	Type	Description
provider	string	Operator of the first NS host, from its registrable domain.
domains	integer	Number of domains.
share_of_ns_domains_pct	numeric	domains / domains with NS x 100, two decimals.

certificate_issuers.csv

The 25 largest certificate issuers, plus Other for the rest.

Column	Type	Description
issuer	string	Issuing organisation of the leaf certificate, merged by family.
certificates	integer	Certificates observed from this issuer (scan rows).
share_pct	numeric	certificates / all certificates observed x 100.

certificate_validity.csv

Valid, invalid and self-signed certificates.

Column	Type	Description
measure	string	valid, invalid, or self_signed (a subset of invalid).
domains	integer	Number of domains.
share_of_with_cert_pct	numeric	domains / domains presenting a certificate x 100, four decimals.

web_server_header.csv

The 25 most common Server header values. Values ending .compute.internal are the backend hostnames of one domain-forwarding service; read them as one service.

Column	Type	Description
server	string	Server header value grouped by family, version removed, as observed.
domains	integer	Number of domains.
share_of_server_present_pct	numeric	domains / domains sending a Server header x 100, two decimals.

x_powered_by_header.csv

The 20 most common X-Powered-By values.

Column	Type	Description
value	string	X-Powered-By value as sent, version included.
domains	integer	Number of domains.

population.csv

Every stage from names sourced to domains graded, for this edition.

Column	Type	Description
measure	string	Stage: names_sourced, wildcard_phantoms_removed, idna_duplicates_removed, wildcard_tld_names_added, names_scanned, nxdomain_not_graded, names_that_exist, the four dispositions, and graded.
names	integer	Names at that stage.

Method notes for this edition

Compared with the previous edition, 2026-09 (as of 5 September 2026):

• The same 12 files carry the same field layout. checks.json adds 6 measures to its all-TLD block: cert_expiry, cert_sigalg, cert_key, https_redirect, ns_diversity, soa.
• Grading uses the same methodology and grade bands, with one engine change. On 15 September 2026 the engine added the RFC 7489 organisational-domain DMARC fallback: when a scanned name below a registrable domain has no _dmarc record of its own, the engine reads the record of its organisational domain. In this edition the fallback supplied the DMARC policy of 5,211 of the 332,492,319 graded domains (0.0016%). For a name that is itself a registrable domain the lookup is unchanged.
• Both editions grade on the names that exist. 2026-09 removed non-existent names with a pre-scan existence check (154,021,121 names). 2026-10 removed them by the scan's own NXDOMAIN answer (271,599,411 names). Existing-but-dead names, graded F in both editions: 1,389,076 in 2026-09 and 2,887,602 in 2026-10.
• 2026-10 takes CommonCrawl from all 54 webgraph releases, and breaks out 520 TLDs against 497 in 2026-09.
• The 2026-10 files list each internationalised (xn--) TLD under its own key in every per-TLD map. The 2026-09 checks.json pooled them under one xn key.
• The 2026-10 checks.json _meta.graded_note supersedes the 2026-09 one. Graded means the graded disposition plus the dead names that exist, with NXDOMAIN names excluded.
• Each edition is a snapshot of its own scanned population. Shares from two editions are not differenced as a trend.

Licence

Creative Commons Attribution 4.0 International (CC BY 4.0), https://creativecommons.org/licenses/by/4.0/. Copyright (C) 2026 Defaults Exposed FZ-LLC. Reuse for any purpose, commercial work included, with attribution to Defaults Exposed FZ-LLC and the concept DOI 10.5281/zenodo.22690388.

Load a table

Each table is its own configuration; grades_by_tld opens by default. Each edition is tagged with its edition id, so a pinned revision always returns the same figures.

from datasets import load_dataset

grades = load_dataset("defaultsexposed/domain-security-census", "grades_by_tld", split="census", revision="2026-10")