Skip to content

Recently Isolated Domain

A domain whose harmonic centrality places it in the deep tail (very few distinct sources link to it) while PageRank is also in the mid-to-tail range. Two independent link-graph metrics simultaneously indicate isolation, which is typical for throwaway phishing infrastructure, orphan C2 nodes, or disposable hosting flips.

Also surfaces as:
Persistently Isolated Domain
Analyst Q&A

What this signal answers

The questions every analyst asks before acting on this signal.

What it detects01

What does 'Suspected Narrow Reach Pattern' mean in SOC analysis?

A domain whose harmonic centrality places it in the deep tail (very few distinct sources link to it) while PageRank is also in the mid-to-tail range. Two independent link-graph metrics simultaneously indicate isolation, which is typical for throwaway phishing infrastructure, orphan C2 nodes, or disposable hosting flips.

Why it matters02

What is the difference between Transient and Persistent narrow reach?

Transient indicates the isolation pattern is observed only in the latest period, could be a freshly registered domain or a temporary visibility dip. Persistent means the domain has stayed isolated across at least persistent_min_count consecutive periods, supporting the interpretation of intentionally hidden or orphan infrastructure.

How analysts use it03

Where is the intensity (both-ranks-deep) information reported?

When both harmonic_pos and pr_pos are simultaneously in deep-tail territory (default >=100M each), the label payload's details field includes a both_deep flag and the underlying rank values. Downstream consumers can use this for higher-confidence isolation interpretation independent of the persistence sublabel.

Worth knowing04

How is narrow_reach different from bad_popularity?

bad_popularity is driven by PageRank only and fires when a domain sits in the PR deep tail regardless of harmonic centrality. narrow_reach fires when harmonic is in deep tail AND PR is in mid-to-tail range, covering the PR 10M-50M population that bad_popularity misses. The guard ensures the two labels never co-fire, eliminating amplification double counting.

Worth knowing05

Why can a brand-new legitimate domain trigger this label?

Yes, newly registered legitimate domains may not yet have any inbound links and can fall into the isolated zone, typically with the transient sublabel. The scoring amplification mechanism is designed for exactly this: with no malicious co-signals present, the base impact remains very low (0.1) and the label stays informational. Persistent narrow reach combined with malicious co-signals escalates the score significantly.

Often paired with:alert_only
Evidence shape

What Recently Isolated Domain looks like in the response

A realistic shape of the entry the engine appends to the risk_signals array when this signal fires. Evidence keys are subgroup-default; per-signal overrides documented in the API reference.

json
{
  "slug": "narrow_reach",
  "name": "Recently Isolated Domain",
  "group": "popularity_and_backlink",
  "subgroup": "popularity_score",
  "direction": "malicious",
  "evidence": {
    "tranco_rank": 482103,
    "tranco_rank_band": "100k-1m",
    "organic_traffic_band": "low",
    "regional_rank_top_country": "US"
  }
}
See in API reference
Siblings

Peers in Popularity Score

9 other signals share the same subgroup. They detect related behaviors and often co-fire on the same IOC.

Emerging Link Farm Pattern

A domain that is reachable from many distinct sources (high harmonic centrality) yet none of those sources carry meaning

Malicious
Newly Broadly Reachable

It represents a domain that receives links from a large number of distinct sources (high harmonic centrality) but whose

Safe
Newly Niche-Authoritative

A domain that is referenced by a small number of sources (mid or tail harmonic centrality) but those sources carry stron

Safe
High-Reputation Domain

It refers to globally recognized, high-traffic domains with verified ownership and a long operational history, such as g

SafeCore
Show 5 more in Popularity Score

Start free, in five minutes

1,000 verdicts/month on the free tier, LLM layer included. No card required.