Pollution reports
The published IPI 2.0 months, for any member's API key: each month's pollution index and totals, its economy scores, and four gzipped JSON files served from disk.
GET /api/pollution/reports published months, newest first
GET /api/pollution/reports/{yearMonth} one month, with each economy's score
GET /api/pollution/reports/{yearMonth}/economy-scores
GET /api/pollution/reports/{yearMonth}/asn-pollution
GET /api/pollution/reports/{yearMonth}/ip-points
GET /api/pollution/reports/{yearMonth}/ip-attrs
{yearMonth} is YYYY-MM; anything else is a 422. A well-formed month with no report is a 400 with
detail Not found. A month is published once its economy scores are exported.
Currents' server reads these routes with its own key and publishes what they return, so the JSON
keys inside the files and the column ordering in the two columnar files do not change. The keys keep
the word "country" for that reason, though the file is economy-scores.
The pipeline behind them: any member may read GET /api/pollution/pipeline, each month's place in it.
GET /api/pollution/pipeline/advancing and POST /api/pollution/reports/{yearMonth}/regenerate and
/export are admin-only.
Transport
The body is pre-gzipped on disk and served as-is with Content-Encoding: gzip. The response
compression middleware carves out these files precisely so the payload is not gzipped a second time,
which would leave a client unable to read it.
The files are large and change only when a month is regenerated, so they revalidate rather than expire:
Cache-Control: public, max-age=0, must-revalidate
ETag: "18d159daa064c30a-19e"
Last-Modified: Wed, 02 Sep 2026 00:35:22 GMT
Send either validator back and an unchanged file costs a 304 with no body. Do that rather than
re-downloading: ip-points for a busy month runs to megabytes, and the whole set since 2018 is
several hundred.
curl -H 'If-None-Match: "18d159daa064c30a-19e"' \
https://gcaaide.org/api/pollution/reports/2026-01/economy-scores
economy-scores
Per-economy pollution, plus a corpus-wide summary.
{
"summary": {
"totalCountries": 3, "totalDistinctAsns": 3, "totalDistinctIps": 80,
"totalHits": 2240, "totalScans": 560, "totalScouts": 560,
"totalIntrusions": 1120, "avgPollutionPct": 85.13666666666667
},
"countries": [
{
"reportingMonth": "2026-01-01T00:00:00Z",
"countryCode": "NL", "countryCode3": "NLD",
"nIPsCountry": 28, "HCountry": 0.5796875, "MpCountry": 16.23125,
"pollutionScore": 0.8665997505835371, "pollutionScorePct": 86.66,
"countryAddressSpace": 512,
"sumScanning": 168, "sumScouting": 224, "sumIntrusion": 392,
"totalHits": 784, "asnCount": 1
}
]
}
Ordered by pollution score, highest first.
asn-pollution
Three fields and no more. It is the address-space denominator, not a scores file.
[ { "asn": "14061", "countryCode2": "NL", "asnAddressSpace": 512 } ]
asn is a string, not a number: it is a dictionary key on the far side.
ip-points and ip-attrs
Two columnar files, split so a map can plot before the attributes finish loading. Both are parallel
arrays of count values, and row i in either file is the same address, which is the whole
reason they can be separate.
ip-points carries what is needed to plot and score a point:
{
"count": 80,
"lat": [52.379, ...], "lon": [4.899, ...],
"scan": [0, ...], "scout": [28, ...], "intr": [0, ...],
"days": [14, ...], "rp100": [100, ...]
}
daysis how many days of the month the address was seen.rp100is the AbuseIPDB reputation scaled by 100. An address with no reputation on file scores 100, so 100 means "nothing known", not "known clean". Reputations cover the reporting month only; months looked up before that change covered 240 days, so their scores and botnet shares read higher.
ip-attrs carries the filtering and tooltip fields, dictionary-encoded:
{
"count": 80,
"asnDict": ["14061", "396982", "45102"], "asn": [0, 0, 1, ...],
"ccDict": ["NL", "US", "SG"], "ccNames": ["Netherlands", ...], "cc": [0, 0, 1, ...],
"usageDict": [""], "usage": [0, ...],
"tldDict": [""], "tld": [0, ...],
"botnet": "AAAAAAAA..."
}
Each *Dict lists the distinct values and its partner array holds one index per row. botnet is a
base64 byte array, one byte per row.
usage and tld come from AbuseIPDB, so an address it was not asked about carries the empty
dictionary entry.
Addresses with no geo are absent
An address whose geo lookup failed is not in these files at all. That matches what AIDE published: the pipeline drops a row with no coordinates rather than plotting it at 0,0, which is a valid point in the Gulf of Guinea that MaxMind never actually returns.