You buy a proxy, check the address, and every score comes back green. Then an account registered through it is suspended within a day, and the IP still scores green afterwards.
The address was never the unit being judged. The prefix was.
Three scopes that get confused
An IP reputation lookup answers one narrow question about a single 32-bit number. What a target site enforces is usually decided at a coarser scope, because that is where abuse is cheap to detect and expensive to launder.
| Scope | Unit | Who decides | What a block means |
|---|---|---|---|
| Address | One IP | That site's own rules | This endpoint is burned |
| Prefix | /24 or the RIR allocation | Abuse feeds, CDN rule sets | Everyone inside inherits it |
| ASN | Entire network | Onboarding and KYC teams | The whole provider is filtered |
A green score is sampled at the first scope. Enforcement frequently happens at the second. That gap is the whole problem.
Why one bad tenant lands on you
Three paths carry the penalty, and none of them requires you to have done anything.
Feed replication. An abuse feed records one address for sending junk. Consumers of that feed do not always apply it as a single host — many apply it to the covering /24, because spam operations rotate addresses inside a prefix they control. One entry becomes 256.
Registration clustering. RDAP and WHOIS describe the block as one allocation with one organisation, one abuse contact, one netname. Risk engines read that as a unit and inherit the allocation's history as context for each address inside it.
Observed density. No blocklist is needed. The target counts: four hundred new accounts, two hundred and fifty-six addresses, one TLS fingerprint, one inter-arrival distribution. The neighbourhood is guilty by statistics rather than by list.
The third path is the one that catches people with genuinely clean addresses. Nothing to appeal, no record to remove.
Shared, segment-dedicated, full-dedicated
Vendors use these words loosely. The distinction that matters is how many strangers sit inside your /24.
| Plan style | Addresses you use | Strangers in your /24 | What you inherit |
|---|---|---|---|
| Shared pool | A few, rotating | Up to 253 | Every tenant's history and behaviour |
| Segment dedicated | A slice of a prefix | Usually 2 to 4 tenants | Their history, diluted |
| Full dedicated | A /24 or larger | None, unless self-inflicted | Only your own patterns |
"One dedicated IP" sits in the first row. A dedicated address is not a dedicated neighbourhood, and that is where most expectations break.
Counting your neighbours
Reverse DNS is a passive way to see who else is announced in the block: hosting providers publish PTR records in bulk, residential blocks mostly do not. No packets reach the addresses themselves.
import ipaddress, socket
from collections import Counter
from concurrent.futures import ThreadPoolExecutor
def ptr(ip):
try:
return str(ip), socket.gethostbyaddr(str(ip))[0]
except (socket.herror, socket.gaierror):
return str(ip), None
addr = ipaddress.ip_address("203.0.113.57")
net = ipaddress.ip_network(f"{addr}/24", strict=False)
sample = list(net.hosts())[:64] # never sweep a whole /24
with ThreadPoolExecutor(max_workers=16) as pool:
rows = list(pool.map(ptr, sample))
named = [r for r in rows if r[1]]
print(f"{len(named)}/{len(rows)} sampled addresses publish PTR")
for ip, name in named[:10]:
print(ip, "->", name)
# Last two labels usually identify the operator, not the customer.
tails = Counter(".".join(n.split(".")[-2:]) for _, n in named)
print(tails.most_common(5))
Two numbers matter in the output. The share that resolves tells you whether this is a hosting block or a residential one. The distinct tails tell you how many operators are diluted into one prefix.
Reading the count correctly
High density is not automatically bad. A hosting /24 with two hundred PTR records is a normal web host and is treated as one. A residential /24 with two hundred PTR records is a farm, and feeds learn it fast.
Low density is not automatically good either. A /24 where nothing resolves may be freshly allocated, or may be a sinkhole operated by someone who never bothered with reverse DNS. Ask for the allocation date from RDAP when the picture is ambiguous.
What you are looking for is one specific combination: many operators inside one prefix, recent allocation, and a residential or mixed-use label. That profile is where collateral damage concentrates.
Common mistakes
- Trusting an address-level score as a verdict on the neighbourhood. It is one sample of 256.
- Buying "dedicated IP" and expecting a dedicated prefix. You have one address and up to 253 strangers.
- Treating PTR density as a purity signal in both directions. Direction depends on the hosting label.
- Testing one target and concluding the block is fine. Different consumers apply different granularities.
- Switching vendors that resell the same upstream ASN — same neighbourhood, same result, new invoice.
FAQ
Does a clean score mean I will not get blocked?
No. It means the single address is not on the lists that checker reads. Prefix-level feeds, behavioural density, and site-specific rules are outside its scope.
Can I clean up a dirty prefix?
Not the other tenants — you have no authority over them. You change scope: move to a segment or a full /24, or reduce the behaviour that made the block statistically attractive.
Why does the same IP work on one site and fail on another?
Different feeds with different granularity. One site applies the entry to a single host, the next applies it to the covering /24 and you inherit it without appearing anywhere.
Are residential IPs safer because they look human?
Not by themselves. Residential prefixes are dense by design, and carrier-grade NAT hides the offender inside a shared address, so feeds list those blocks quickly. The label helps; the density still counts.
If you want to see your own neighbourhood before a target does, start by reading the allocation — the Chinese walkthrough of neighbour effect and how to check it is at socks5ip.com.cn. For the address-level checks that the prefix view sits on top of, use the free tools at socks5ip.com.cn and compare the two answers; when they disagree, the prefix is the one that gets enforced.
Top comments (0)