How AbuseTrack risk scoring works
Every IP address on AbuseTrack carries a community risk score from 0 to 100. It answers one question: how confident should you be, right now, that this address is abusive? This page documents exactly how the number is computed, so you can interpret it and choose your own thresholds.
The 0-100 scale
The score is a single integer between 0 and 100, computed from all approved reports for an address. It is the sum of three capped components, rounded and clamped to 100. The largest component is corroboration - how many independent reporters flagged the address - because agreement between unrelated reporters is the strongest signal that abuse is real rather than a one-off or a grudge.
The formula
Each component is capped, so no single factor can dominate. Weights come from the decay function described below.
| Component | Calculation | Maximum |
|---|---|---|
| Corroboration | weighted distinct reporters × 18 | 50 |
| Report volume | weighted reports × 6 | 30 |
| Threat-type variety | (distinct threat types − 1) × 6 | 20 |
score = round(min(100, corroboration + volume + variety))Corroboration is capped at half the scale, so one reporter on their own can never push an address to High.Corroboration (up to 50 points)
Each distinct reporter counts once, however often they report the same address. Three recent, independent reporters are enough to reach the cap.
Report volume (up to 30 points)
The decay-weighted number of reports. More reports raise the score, but the cap means a flood of duplicates cannot dominate the result.
Threat-type variety (up to 20 points)
An address reported for more than one kind of abuse - for example brute force and port scanning - gains 6 points per additional threat type. A single threat type adds nothing here.
Recency and decay
Threats are not permanent. Every report is weighted by its age with exponential decay and a half-life of 21 days. An address that stops attracting reports cools down on its own, and a high score reflects recent activity rather than old history. Nobody has to expire old reports by hand.
| Report age | Weight |
|---|---|
| Today | 100% |
| 21 days | 50% |
| 42 days | 25% |
| 63 days | 12.5% |
| 90 days | ≈ 5% |
weight = 0.5 ^ (age_in_days / 21)
How reporters are counted
Corroboration is measured per reporter, not per report. Each reporter is weighted by the age of their most recent report on that address, so someone who flagged it months ago contributes less than someone who flagged it yesterday.
All anonymous and guest reports collapse into a single reporter. They can add supporting weight, but submitting many unauthenticated reports cannot manufacture corroboration.
Worked examples
Three cases, assuming all reports were made today (weight 1):
| Situation | Corroboration | Volume | Variety | Score |
|---|---|---|---|---|
| 3 reporters, 5 reports, 2 threat types | 50 | 30 | 6 | 86 High |
| 1 reporter, 20 reports, 1 threat type | 18 | 30 | 0 | 48 Medium |
| 1 reporter, 1 report | 18 | 6 | 0 | 24 Low |
Low / Medium / High
The numeric score maps to three levels:
| Level | Score | Meaning |
|---|---|---|
| Low | 0-24 | Little or no corroborated, recent evidence of abuse. |
| Medium | 25-59 | Some recent, corroborated reports. Treat with caution. |
| High | 60-100 | Strong, recent corroboration from multiple reporters. |
Interpreting the score
Use the score as one input, not a verdict. A High score is a strong signal to block or rate-limit; Medium is a reason to look closer; Low means there is not yet enough recent, corroborated evidence - not a guarantee of safety.
When you pull the blocklist you choose the threshold yourself: the score and reporters parameters of /api/v1/blocklist.txt set the minimum score and the minimum number of distinct reporters.