MCP documentation menu

Detecting malicious bulk registrations

A guide to triaging bulk registration campaigns. datapulse_dns_regclusters surfaces registrars with unusually many registrations on a UTC day; similarity search, scraping and history then test whether a cluster is a campaign. The tool’s parameters, paging and response fields are in datapulse_help(topic="infrastructure").

Everything below is a triage heuristic. Registrar, TLD, naming and timing patterns are reasons to look closer; a finding needs a second, independent signal.

Red flags

Registrar

SignalWhat to look for
Abuse handlingA registrar whose abuse desk your own takedown records show as slow to act; this is the analyst’s own record, not a property of any given domain
Unfamiliar registrar, high volumeA registrar not usually in the top clusters appearing with hundreds of registrations a day
Volume spikeA registrar’s count far above its own count on adjacent days
Recurring appearanceThe same registrar in the top clusters on several consecutive days

The largest retail registrars register tens of thousands of names a day; their volume is normal traffic and says nothing in either direction about a particular cluster.

Domain names

sample_domains in summary mode shows a cluster’s names, and detail mode the full list:

PatternExampleCommonly
Random 5–8 charactersxagyd.app, kqmvf.netAlgorithmically generated (DGA)
Number strings0551866.top, 8837421.icuSpam, parking, SEO
Brand misspellingamaz0n-login.com, paypai-secure.netPhishing
Brand + random suffixapple-verify-8x3k.comPhishing
Gambling keywordsplaywin222.net, bet-bonus99.topIllegal gambling
Crypto + brandbinace-wallet.com, metamask-claim.xyzCrypto scams
Pharma keywordscheap-viagra-rx.sbs, pharmacy-online.clickPharma spam

Brand-carrying names are triaged as in datapulse_help(topic="lookalike_domains").

TLDs

WeightTLDs
High abuse rates in public abuse data.top, .icu, .sbs, .cyou, .xyz, .click, .vip
Elevated.site, .online, .fun, .store, .club, .buzz, .surf
Context-dependent.app, .dev, .io: mostly legitimate, also used with brand names

Registration timing

Detail mode gives each domain’s registration_date:

PatternSignificance
Many registrations within seconds of each otherAutomated tooling
Registrations concentrated in a 1–2 hour windowA bulk purchase batch
All at the same minuteOne API call or bulk provisioning
Spread evenly across the dayConsistent with ordinary retail traffic

A recent day’s counts keep growing for up to about three weeks as registration data arrives, so a low count on a recent day is not a drop in activity.

Investigation workflow

1. Surface clusters

datapulse_dns_regclusters(date="2025-02-09", min_cluster_size=50)

Each row carries registrar_id, registrar_id_scheme (iana for an IANA-assigned ID, registry for a ccTLD registry’s own number), registrar_name when the ID is an IANA one, domain_count and up to 10 sample_domains.

2. Drill into a registrar

datapulse_dns_regclusters(date="2025-02-09", registrar_id=3222)

Detail mode lists that registrar’s registrations for the day, 200 per page by default, with total and next_offset. The first page usually shows the naming pattern, the TLD mix and the timing.

3. Infrastructure overlap

Two or three domains from the cluster as similarity seeds:

datapulse_dns_dptechsim(domain="suspicious-domain.top", similarity=0.90)

0.90 widens past the 0.95 default to names sharing major components. Shared rare elements across unrelated-looking names (the same uncommon nameserver ASN, the same mail host, the same registrar IANA ID) strengthen the case for coordination. Overlap on a large CDN or cloud network is common to millions of names and is not evidence either way.

4. Content

datapulse_scrape_submit(url="http://suspicious-domain.top", wait=true)

Parked or blank pages, redirect chains (redirect_chain), brand login clones, SEO doorway pages and crypto landing pages are the common findings.

A scrape that fails usually means no web address is published:

datapulse_live_dns(domain="suspicious-domain.top")

No A/AAAA records explains a failed scrape, though a scrape can also fail when addresses exist (blocked, timed out, no HTTP service), and the name may still answer for MX, TXT or NS. Names registered without web addresses are common in bulk registrations: some are staged for later use, others are held defensively or for resale. On its own that is a triage cue.

5. Timeline

datapulse_dns_dphistory(domain="suspicious-domain.top")
datapulse_live_rdap(query="suspicious-domain.top")

The hosting history records nameserver and MX hostnames with ASNs (no operator names, no countries) and the web-hosting ASN, mostly at month granularity. A first listing (a) that is recent, or a listing that appeared and was de-listed (d) quickly, fits a burned or short-lived domain. It is too coarse to show fast address rotation within a day. Listing events are not registry events: datapulse_live_rdap gives the registration date and status. See datapulse_help(topic="history").

Example: one detail page

A detail-mode call for registrar 3222 on 2025-02-09 (rows as returned on 2026-10-19):

datapulse_dns_regclusters(date="2025-02-09", registrar_id=3222)
→ registrar_id 3222, registrar_id_scheme "iana", registrar_name "Domainipr Limited"
→ total 349
→ domains include 003370.com, 0755bzh.com, 09xa.com, 0bxx.com, 107758.com
→ registration_date values between 06:18 and 19:34 UTC

Against the red flags: short number-and-letter labels on .com fit the “number strings” pattern; the registrations span most of a business day rather than one burst. That is a reason to run steps 3–5 on a few of the names, not a conclusion about the registrar or the batch.

Legitimate versus malicious

FactorPoints away from abusePoints toward abuse
VolumeConsistent day to day for that registrarA sudden spike
Domain namesReal words, brandableRandom strings, brand impersonation
TLDs.com, .net, .org.top, .icu, .sbs, .cyou
TimingSpread across the dayA concentrated burst
InfrastructureEach name on its own established stackMany unrelated-looking names on one uncommon stack
Web contentReal websitesParked, redirecting, phishing
RegistrationOlder names, deeper locksNew names, default locks

Registrar size and large-CDN hosting appear in neither column: they describe most of the internet.

Triage checklist

  • Registrar usually in the top clusters, or new to them?
  • Sample names human-generated?
  • TLDs high-abuse?
  • Volume normal for this registrar on adjacent days?
  • Infrastructure shared across samples (dptechsim on 2–3)?
  • Content live, and what is it (scrape 1–2)?

Three or more raised flags make a cluster worth a full investigation and, when the content confirms abuse, a report to the registrar’s abuse contact (abuse_email in datapulse_live_rdap).

Generated from the live server (DataPulse MCP v0.9-341-g27163c2) on October 20, 2026. Your AI assistant reads this page by calling datapulse_help(topic="malicious_bulk_regs").