Skip to main content

haversine Search Operator

The haversine operator returns the distance between latitude and longitude values of two coordinates in kilometers. Coordinates need to be positive or negative values based on being north/south or east/west, instead of using the terms N/S, E/W.


haversine(<latitude1>, <longitude1>, <latitude2>, <longitude2>) as <field>


| haversine(39.04380, -77.48790, 45.73723, -119.81143) as distanceKMs

This returns a field named distanceKMs with the value '3,512.71000'.

Return value in miles

To convert kilometers (KM) to miles you can divide the KM value by 1.609344.

| haversine(39.04380, -77.48790, 45.73723, -119.81143)/1.609344 as distanceMiles

Landspeed violation example

You can use the following query detect landspeed violations in AWS CloudTrail with haversine:

| json "userIdentity.userName" as user nodrop
| json "sourceIPAddress" as ip nodrop
//| filter and sort the data
| where user matches {{actor}}
| where isPublicIP(ip)
| min(_messagetime) AS login_time BY user, ip
| sort BY user, +login_time
//| Next, find the previous IP address where each user logged in from, and the previous login time.
| ipv4ToNumber(ip) AS ip_decimal
| backshift ip_decimal BY user
| backshift login_time AS previous_login
| where !(isNull(_backshift))
//| Next, we’ll convert the IP addresses that are in a decimal format to the standard IP address using octets.
| toInt(floor(_backshift/pow(256, 3))) AS octet1
| toInt(floor((_backshift-octet1*pow(256, 3))/pow(256, 2))) AS octet2
| toInt(floor((_backshift-(octet1*pow(256, 3)+octet2*pow(256, 2)))/256)) AS octet3
| toInt(_backshift-(octet1*pow(256, 3)+octet2*pow(256, 2)+octet3*256)) AS octet4
| concat(octet1,".",octet2,".",octet3,".",octet4) AS previous_ip
//| Now that we have two different IP addresses, we can use geo lookup on both to find our where each occurred.
| lookup latitude AS lat1, longitude AS long1, country_name AS country_name1 FROM geo://location ON ip
| lookup latitude AS lat2, longitude AS long2, country_name AS country_name2 FROM geo://location ON ip=previous_ip
//| Now that we have two geolocations from two successive logins, we can calculate the distance between them using the Haversine function
| haversine(lat1, long1, lat2, long2) AS distance_kms
//| Calculate the speed someone would need to travel that distance in the time between the two logins
| (login_time - previous_login)/3600000 AS login_time_delta_hrs
| distance_kms/login_time_delta_hrs AS apparent_velocity_kph
| where apparent_velocity_kph > 0
//| Flag certain speeds as suspicious.
| 500 AS suspicious_speed
| where apparent_velocity_kph > suspicious_speed
//| add some formatting to clean up the results and make them more human-readable. Add these lines to your query:
| concat(ip,", ",previous_ip) AS ip_addresses
| if(country_name1 <> country_name2, concat(country_name1, ", ", country_name2), country_name1) AS countries
| fields user, ip_addresses, countries, distance_kms, login_time_delta_hrs, apparent_velocity_kph
| where !isNull(user)
| where apparent_velocity_kph != "Infinity"
| sort by apparent_velocity_kph
Privacy Statement
Terms of Use

Copyright © 2024 by Sumo Logic, Inc.