*Published: 10/2/2026*
> Note: Still a running WIP!
> **High-level Overview:** Phishing operators buy domains the way they buy anything else they plan to throw away, and they buy the cheapest one that will last long enough to do the job, and that one decision explains most of the TLD abuse data. A handful of new generic TLDs sell their first year for under two dollars and charge seven to seventeen times that to renew, so the attacker registers, burns, and replaces without ever seeing the second bill. Malware operators price things differently and lean on stolen reputation, and anyone who needs to stay up longer moves onto developer platforms, where blocking the parent domain would also block legitimate projects. The block list comes first, and the rest of the piece walks through the economics that produced it.
## Why block a whole TLD
Blocking one domain at a time means chasing domains that are registered, used, and dropped faster than reputation feeds can list them. In six months, new .bond registrations came to 98.6% of the zone's total size, and attackers burn through those domains as each one is detected and taken down, long before anyone would pay to renew them [^1].
At the TLD, one rule at the DNS resolver, the web proxy, or the mail filter covers every domain in the zone, including the ones that have not been registered yet, and it costs almost nothing when the zone carries little legitimate use. That cost decides where each zone lands, so the list below splits them by how much real use they carry:
- Block the whole zone
- Block only newly registered domains
- Block by hostname
- Leave it open
## What to block
A TLD lands in Tier 1A when it is at least four times worse than .com on two or more of these four public measures:
- The share of its 2025 registrations later blocklisted as malicious
- The share of the zone reported for phishing
- The share of the zone on Spamhaus lists
- Malware domains for every 100,000 in the zone
A measure only counts when the zone holds at least 5,000 domains and at least 50 abuse domains sit behind the figure, or 25 malware domains for the 90-day abuse.ch window, so a tiny zone with a handful of bad domains cannot trip the rule. Every TLD in the four datasets was run through that test, but a block also has to avoid breaking anything you need, so where a zone lands depends on two conditions:
- A zone is blocked outright, in Tier 1A or 1B, when it meets that abuse rule, or is a known junk zone, and has almost no legitimate use
- A zone goes to Tier 2, where only newly registered domains get blocked, when it clears the abuse bar but carries real services, the way .run carries the any.run sandbox, or when it clears the test on a single measure
As a rough gauge of legitimate use, the piece counts how many of each zone's sites appear in the Tranco top million [^2]. Tier 1B holds the junk zones blocked for structural reasons the ratios cannot capture, and those reasons are covered under "Where defenders cut in".
| Tier | Action | TLDs |
|---|---|---|
| 1A | Block the TLD | .top .xin .cfd .cyou .icu .sbs .bond .lat .mom .bid .mobi .life .garden .qpon .loan .help .rest .win .ink .pizza |
| 1B | Block the TLD | .zip .mov .su .tk .ml .ga .cf .gq .homes .boats .autos .hair .cam .best .wang |
| 2 | Block domains under 30 days old | .live .wiki .run .cc .pro .vip .click .shop .xyz .info .online .digital .cn .lol .site .store .space .fun .asia .biz .finance .world .pics .fit .support .club .buzz .one .today .quest .bar .fyi .monster .baby .network .guru .forum .food .bet .casino .locker .lgbt .town .gdn .pictures .auction .pink .reviews .claims .pet .forsale .date .loans |
| 3 | Leave the TLD open, block by hostname | .dev .app .io .me .co .ai .page .sh .gg .tv .link |
| 4 | Leave open | .com .net .org .edu .gov .mil .int .us .uk .ca .au .de .fr .eu |
| Watch | Monitor, or block new domains if you have no business there | .ru .jp .my .ng .ua .pk |
![[tld_v9_tier_map.png]]
Two of the four datasets publish every TLD, with the Interisle phishing table covering 233 TLDs including country codes and the abuse.ch data covering all 362 TLDs that carried malware in the window. The other two only publish their worst performers, with Interisle's registration study naming only its worst gTLDs and Spamhaus printing top-20 tables, so a zone missing from those lists may simply be unmeasured. The platform hostnames in Tier 3 get their own block list under "Why the platforms still win".
## Where the phishing actually sits
Ranked by raw count, .com is the worst TLD on the internet, with Interisle counting 869,180 domains reported for phishing in .com over a single year, more than double .top in second place, yet that number mostly reflects how big .com is, since it works out to about 1 in 190 of the zone [^3]. Rank the same data by share of the zone, and the picture flips, because in .xin, about 59 of every 100 registered domains were reported for phishing that year, while the next tier down, .win, .icu, and .garden, each sit around 16%. A defender who blocks by volume ends up staring at .com, which can never be blocked, while the zones where most of the domains exist only to phish go unnoticed.
![[tld_v8_volume_vs_rate.png]]
## Ephemeral domains
From the attacker's side, a phishing domain is a consumable with a short shelf life. It gets reported, it lands on a blocklist and it stops working, so the only question worth asking at purchase is what each one costs. In the price lists, every TLD where more than 1 domain in 10 was reported for phishing sells its first year for under $5, a group that includes .pro alongside .xin, .win, .icu, .garden, .cfd, and .help [^4]. Twelve zones go further and sell the first year for $1.54 to $1.80, against $11.08 for .com. Interisle saw the same pattern in its 2025 study, where 18 of the 25 new TLDs with the highest phishing rates offered non-promotional registrations under $2 at some registrar [^5].
At $1.54 a domain, a phishing crew can buy seven of them for the price of one .com. When each domain only has to survive a few days, the cheap zone wins on cost per working day, and there is no reason to pay for reputation that the page will burn through anyway.
![[tld_v9_price_vs_phishing.png]]
## Nobody pays for year two
The pricing on these zones is lopsided, with those twelve zones charging 7 to 17 times the first-year price to renew, so .cfd costs $1.80 to register and $15.96 to keep [^4]. For a legitimate owner, the renewal is the real price of the domain, while for an attacker it rarely comes up, because, as Spamhaus notes in its reputation updates, attackers use bulk-registered, disposable domains and burn through them as each one is detected and taken down [^1].
![[tld_v5_first_year_vs_renewal.png]]
The turnover shows up in the zone data, where new .bond registrations came to 98.6% of the zone's total size in six months, against the 10 to 20% of new domains that Spamhaus already considers unusually high [^1]. The people buying are the attackers themselves, since 94% to 99% of the phishing domains in .bond, .cyou, .xin, .cfd, .icu, and .sbs were registered for the purpose, compared with 66% in .com and 39% in .cn, two zones that still carry a large number of hacked legitimate sites [^3]. In the cheap zones, the attacker is the customer, paying the registry directly for every domain. In .com, the attacker is far more often a squatter on someone else's site, paying nothing for the domain and getting its age and reputation for free.
The blocklists do catch up eventually, and about a third of the 5.2M .top domains registered in 2025, 1.79M of them, were blocklisted as malicious by April 2026, against 1 in 20 for .com, and the share ran higher still for .xin at 59% and .mobi at 63% [^6]. Spamhaus treats 5% of a zone listed as a red flag, and .cfd sits at 17.5%, .qpon at 12.2%, and .icu at 9.7% [^1].
## A handful of sellers
A short list of companies operates most of these zones, with Shortdot SA running .cfd, .sbs, .bond, .icu, and .cyou and XYZ.COM running .xyz, .lol, .mom, .homes, .lat, .autos, .boats, and .hair [^7]. The concentration carries through to retail, where the registrar NICENIC saw more phishing domains in a year (272,696) than the 237,724 gTLD domains it had under management [^8]. For a defender, this is useful, because a new TLD launched by one of the same registries is worth watching from its first day.
## Phishing and malware shop differently
Malware delivery follows a different set of economics, and the abuse.ch data for July through October 2026 shows it plainly. .com still leads by count with 9,187 malware domains, mostly ClearFake, ClickFix fake-CAPTCHA pages, and unattributed loaders. Sekoia documents both ClearFake and ClickFix lures as JavaScript injected into compromised websites, mostly WordPress, so much of that count sits on hacked sites rather than bought ones [^9][^10]. An old domain carries a benign reputation built over time, which lets it go undetected longer once it turns malicious, and a new two-dollar .click domain has none of that, so loader operators would rather steal an old site than buy a new domain [^11].
When malware operators do buy, they cluster by family, with ACR Stealer leading in .cfd and .cc, ClickFix fake-CAPTCHA pages making up most of .icu and nearly all of .life, and .click split mostly between the Remus and RevStealer infostealers and ClickFix [^12]. Sized against the zone, .garden and .mom carry the most malware of the TLDs with an Interisle zone size, at 135 and 49 malware domains for every 100,000 registered, while .com stays under 6 [^13]. Almost all of the .garden figure comes from a single ClearFake cluster.
.xyz has the second most malware domains of any TLD at 1,544, but 995 of them carry the same MetaStealer tag, so raw counts mislead in both directions. .xin had no malware domains in the window and .mobi had two, even though most of their new registrations were blocklisted, because their abuse is phishing and abuse.ch does not track phishing. .top carries less malware per domain than .com, and its domains show up instead in Smishing Triad's toll and postal SMS lures [^14].
## Why the platforms still win
An attacker who needs a page to stay up for weeks has a better option than any TLD. Phishing on cloud and developer platforms sits on shared hostnames, so the attacker skips the registrar entirely and takes a subdomain on a platform that legitimate projects use every day. Kaspersky notes that this legitimate use complicates bulk blocking, because security teams risk cutting off non-malicious projects [^15]. In its data, pages.dev alone carried 24.9% of the phishing on cloud platforms, and counting only the top ten platforms, at least 71% of it sat under .dev, .app, and .io. Those are the same TLDs engineering teams work on every day, and blocking them would break real work while missing the abuse entirely.
![[tld_v9_dev_platforms.png]]
The control belongs on the hostname, so on the platforms in the chart above, plus r2.dev, which ranked third among platforms in the abuse.ch data [^12], block:
- Subdomains first seen in the last 30 days
- Project names carrying brand or security words, in illustrative patterns like captcha-verify-*.vercel.app
- Subdomains no one inside the organization has visited before
## Where defenders cut in
The economics give three answers depending on the zone:
- Where most of the zone is attacker-bought and nothing legitimate depends on it, block the TLD
- Where the zone has heavy abuse alongside millions of real sites, block only the new registrations, since the attacker's domains are almost always young
- Where the abuse rides on platform hostnames, leave the TLD open and block at the hostname
### How the block list compares to .com
The chart below lists, for each blocked zone, only the measures where it runs at least four times the .com figure, next to the .com number for comparison. Only .cfd and .icu are four times worse than .com on all four.
![[tld_v8_why_blocked.png]]
### What the full sweep added
Running every TLD through the same test turned up four more zones over the abuse bar. The small .pizza zone, with about half of its 2025 registrations blocklisted as malicious and almost no popular sites, goes straight to Tier 1A [^6][^2]. .live, .wiki, and .run land in Tier 2 instead, because each carries real services, from the thousands of popular sites on .live to the any.run sandbox on .run.
### Zones over the bar that stay in Tier 2
.cc, .pro, .vip, and .click all clear the abuse bar comfortably, but each carries real use, and .click in particular hosts a lot of ad and redirect infrastructure, so blocking the whole zone breaks ordinary pages.
### Zones to recheck
The test cuts both ways. In the latest quarter, .mobi ran below .com for phishing and .bid ran at about twice .com, so both deserve a second look at the next review [^16].
### Structural blocks (1B)
Tier 1B rests on structure rather than ratios:
- .zip and .mov are real TLDs run by Charleston Road Registry, Google's registry operator, and they get blocked because mail, messaging, and forum apps can turn a typed file name like invoices.zip into a clickable link [^17]
- Quad9 blocked 195M unique .su names, more than in .com, each drawing only about three lookups, which is the pattern malware leaves when it generates throwaway domain names [^18]
- The old Freenom zones (.tk, .ml, .ga, .cf, .gq) were free to register until Freenom stopped registrations in January 2023, and they have little legitimate use left [^19]
### Small junk zones (1B)
.homes, .boats, .autos, .hair, .cam, .best, and .wang are blocked because there is almost nothing legitimate in them to break. Between 2% and 7% of their domains were reported for phishing, against about 0.5% for .com [^3], and none of them carries a meaningful share of the web's popular sites [^2].
### Why .shop, .xyz, and .info stay in Tier 2
These zones have plenty of abuse, but each also hosts thousands of popular sites, and none is four times worse than .com on more than one measure.
### The watch list
The country codes on watch each clear the bar on one measure, and blocking a whole country is a business call more than a security one. .ru stays on watch because its botnet servers swing with single campaigns, dropping 83% in the first half of 2026 after a ClearFake-driven spike [^20].
Before any of this goes in, pull 30 days of DNS logs for the Tier 1 zones and allowlist whatever your environment actually depends on.
## Gaps
Three gaps remain:
- Nobody publishes abuse rates for .dev, .app, and .me registrations themselves, and if those turn out as abused as .xyz, Tier 3 also needs a block on newly registered domains
- .lol sits just under the line on every measure, with its malware rate just short of four times .com, so it stays in Tier 2 for now [^3][^6]
- Every environment depends on a few odd zones, and only its own DNS logs will show which
## Closing
Attackers will keep buying wherever domains are cheapest, so this list will drift. Phishing on .cn rose 270% in a single quarter this year and on .icu 228%, which is why the tiers need a fresh look every quarter [^16]. Price is what moves them, and Freenom is the clearest case, since its free zones once carried 14% of all phishing worldwide and stopped being a source of new phishing domains once Freenom ended registrations in January 2023 [^19].
## References
[^1]: Spamhaus. *Domain Reputation Update: October 2025 to March 2026*. Published 2026-04-15. [spamhaus.org](https://www.spamhaus.org/resource-hub/domain-reputation/domain-reputation-update-oct-2025-mar-2026/)
[^2]: Tranco. *Top 1M list* ID 94GG2, a research-oriented ranking aggregated over 30 days. Retrieved 2026-10-03. [tranco-list.eu](https://tranco-list.eu/list/94GG2/1000000)
[^3]: Interisle CIC. *Phishing Activity in TLDs, 2025-05-01 to 2026-04-30*, with the full per-TLD [CSV](https://www.cybercrimeinfocenter.org/s/Phishing-TLDstats-May2025-Apr2026.csv). [cybercrimeinfocenter.org](https://www.cybercrimeinfocenter.org/phishing-activity-in-tlds-may-april-2026)
[^4]: Porkbun. *Public TLD pricing API*. Retrieved 2026-10-02. [api.porkbun.com](https://api.porkbun.com/api/json/v3/pricing/get)
[^5]: Interisle Consulting Group. *Phishing Landscape 2025* (data 2024-05 to 2025-04). [PDF](https://static1.squarespace.com/static/63dbf2b9075aa2535887e365/t/68dfd315512a675a8fd94ee8/1759499029838/phishinglandscape2025.pdf)
[^6]: Interisle Consulting Group. *Malicious Registrations in the Domain Name Market: An Analysis of 2025 gTLD Registrations*. Published 2026-06. [interisle.net](https://interisle.net/cybercriminaldomaindemand)
[^7]: IANA. *Root Zone Database*. Retrieved 2026-10-02. [iana.org](https://www.iana.org/domains/root/db)
[^8]: Interisle CIC. *Phishing Activity in Domain Registrars, 2025-05 to 2026-04*. [cybercrimeinfocenter.org](https://www.cybercrimeinfocenter.org/phishing-activity-in-registrars-may-april-2026)
[^9]: Sekoia TDR. *ClearFake's New Widespread Variant: Increased Web3 Exploitation for Malware Delivery*. Published 2025-03-18. [sekoia.com](https://www.sekoia.com/blog/clearfakes-new-widespread-variant-increased-web3-exploitation-for-malware-delivery)
[^10]: Sekoia TDR. *Meet IClickFix: a widespread WordPress-targeting framework using the ClickFix tactic*. Published 2026-01-29. [sekoia.com](https://www.sekoia.com/blog/meet-iclickfix-a-widespread-wordpress-targeting-framework-using-the-clickfix-tactic)
[^11]: Palo Alto Networks Unit 42. *Strategically Aged Domain Detection: Using DNS Traffic Trends*. Published 2021-12-29. [unit42.paloaltonetworks.com](https://unit42.paloaltonetworks.com/strategically-aged-domain-detection/)
[^12]: abuse.ch. *ThreatFox IOC database* and *URLhaus database dump*, domain and URL IOCs 2026-07-04 to 2026-10-02. Retrieved 2026-10-02. [threatfox.abuse.ch](https://threatfox.abuse.ch/), [urlhaus.abuse.ch](https://urlhaus.abuse.ch/)
[^13]: Malware rates divide abuse.ch distinct malware domains for 2026-07-04 to 2026-10-02 by Interisle CIC zone sizes for 2025-05 to 2026-04, and the Spamhaus shares for .cc, .live and the national zones divide Spamhaus listings by the same CIC zone sizes, so the periods do not overlap and these figures are indicative.
[^14]: Palo Alto Networks Unit 42. *Smishing Triad global smishing campaign*. Published 2025-10-23. [unit42.paloaltonetworks.com](https://unit42.paloaltonetworks.com/global-smishing-campaign/)
[^15]: Kaspersky Securelist. *Phishers Are Hijacking Legitimate Cloud Infrastructure*. Published 2026-08-04. [securelist.com](https://securelist.com/cloud-platforms-in-phishing/120832/)
[^16]: Interisle CIC. *Phishing Activity: TLDs Quarter over Quarter, 2026-05 to 2026-07*. [cybercrimeinfocenter.org](https://www.cybercrimeinfocenter.org/phishing-activity-quarter-over-quarter-tlds-may-july-2026)
[^17]: Intego, Joshua Long. *Expect .zip and .mov domains to be used in phishing and malware attacks*. Published 2023-05-18. [intego.com](https://www.intego.com/mac-security-blog/expect-zip-and-mov-domains-to-be-used-in-phishing-and-malware-attacks/)
[^18]: Quad9. *H1 2026 Cyber Insights*. Published 2026-08-03. [quad9.net](https://quad9.net/news/blog/trends-h1-2026-cyber-insights/)
[^19]: Interisle Consulting Group. *Phishing Landscape 2023*, Executive Summary. [interisle.net](https://www.interisle.net/PhishingLandscape2023-ExecutiveSummary.pdf)
[^20]: Spamhaus. *Botnet Threat Update: January to June 2026*. Published 2026-07-10. [spamhaus.org](https://www.spamhaus.org/resource-hub/botnet-c-c/botnet-threat-update-january-to-june-2026/)