haraka
haraka-plugin-geoip
provide geographic information about mail senders.
SYNOPSIS
Use MaxMind's GeoIP databases to report geographic information about senders.
This Haraka plugin supports two geoip modules:
| Plugin | geoip module |
|---|---|
| haraka-plugin-geoip | maxmind |
| haraka-plugin-geoip-lite | geoip-lite |
INSTALL (lite)
Install the npm geoip-lite module and update the DBs:
npm install -g geoip-lite
cd node_modules/geoip-lite && npm run-script updatedb license_key=YOUR_LICENSE_KEYINSTALL (maxmind)
The maxmind module requires the manual download of the GeoIP databases. The npm module maxmind-geolite-mirror will download the files for you and also keep them up-to-date if you run it periodically.
mkdir -p /usr/local/share/GeoIP
npm install -g maxmind-geolite-mirror
/usr/local/bin/maxmind-geolite-mirrorDESCRIPTION
GeoIP results are stored in connection.notes.geoip and the haraka-results object at connection.results.get(geoip). The following information is typically available:
continent: NA,
country: US,If the GeoIP city database is available, the following may also be available:
region: CA,
city: San Francisco,
ll: [37.7484, -122.4156],
distance: 1539 // in kilometers
range: [ 3479299040, 3479299071 ],This module also adds entries like this to your logs:
[geoip] US
[geoip] US, WA
[geoip] US, WA, Seattle
[geoip] US, WA, Seattle, 1319kmCalculating the distance requires the public IP of this mail server. See config.distance.
CONFIG
- distance
Performs geodesic distance calculations. Calculates the distance "as the crow flies" from the remote mail server.
This calculation requires a 'from' IP address. This will typically be the
public IP of your mail server. If Haraka is bound to a private IP, net_utils
will attempt to determine your public IP using STUN. If that doesn't work, edit
config/smtp.ini and set public_ip.
- show.city
show city data in logs and headers. City data is less accurate than country.
- show.region in logs and headers. Regional data are US states, Canadian provinces and such.
Set a connection result to true if the distance exceeds this many kilometers.
- too_far=4000
SPAM PREDICTION WITH DISTANCE
Spatio-temporal Network-level Automatic Reputation Engine
"For ham, 90% of the messages travel about 4,000 km or less. On the
other hand, for spam, only 28% of messages stay within this range."Observations in 2014-2016 suggest that geodesic distance continues to be highly correlated with spam.
LIMITATIONS
The distance calculations are more concerned with being fast than accurate. The MaxMind location data is collected from whois and is of limited accuracy. MaxMind offers more accurate data for a fee.
For distance calculations, the earth is considered a perfect sphere. In reality, it is not. Accuracy should be within 1%.