Splunk Search

How to Create "Impossible Travel" Security Monitoring Use Case with pure SPL

marycordova
SplunkTrust
SplunkTrust

I have some reservations about the usefulness of this with so much more usage of IaaS/PaaS/SaaS these days...but since this is non-trivial to produce, I thought I would save everyone the work of developing from scratch if it is something you'd like to monitor. I would also like to note, this has bubbled up activity that was unauthorized/malicious in my experience...so maybe it's not useless...

This uses a macro built on the search string provided by @MuS in this post: https://answers.splunk.com/answers/90694/find-the-distance-between-two-or-more-geolocation-coordinat...

Prerequisites:

  1. normalized fields: user, src_ip
  2. geodistance macro
  3. well filtered base search of 10000 or less events OR
  4. streamstats limits.conf max_stream_window = <adjusted for base search> DO NOT RAISE THIS ARBITRARILY in my environment I have raised this to an upper limit of +10% of my average base events (to include my max) over the past month after doing a LOT of filtering in the base search
  5. filtering clauses to get this down to a manageable number of results
@marycordova
0 Karma
1 Solution

marycordova
SplunkTrust
SplunkTrust
  1. base search: as many indexes/data sources as might contain authentication, authorization, or access data that may be relevant for monitoring potential unauthorized or malicious activity
  2. filter out garbage IP addresses: src_ip!=<rfc 1918 addresses> src_ip!=<whatever other garbage you don't need> maybe you don't really want to try to deal with IPv6 addresses, maybe you have a lookup that you can drop all your known public ip space
  3. filter out garbage user data: user!=*test* user!=<whatever other garbage you don't need>
  4. normalize user: in my case I take whatever user attribute is in the base search, username or email address or whatever, and then query my IAM provider to return a standard value type for my user field, for example email address
  5. dedup for each user: from your base search you want each unique ip address and _time for each individual user, for me this took about 250K events down to about 30K, you would think including _time in the dedup would render the dedup useless but in this case it did not and _time is a critical field, you may also want to include index or sourcetype (whichever is applicable)
  6. prep for streamstats, the "current" data becomes the "destination": dest_time, dest_ip
  7. prep for streamstats, sort your data for each user and _time: streamstats needs to work a single user's data _time sorted so the results aren't broken into different sets (default limit of 10K)
  8. streamstats to create the "source" for each "destination": from the users previous event src_time, src_ip are generated (default limit of 10K)
  9. filter out garbage streamstats data: the first event for every user will have a null src_ip, and events with the same src and dest ip addresses are useless
  10. calculate geo statistics: the geodistance macro will take src_ip and dest_ip as arguments and return the city, region and country for each ip, the distance traveled in km and miles, hours it took to travel, and the kph and mph
  11. additional filtering logic: in my case I dropped all events where mph are less than the speed of a commercial airplane as well as any events where the src and dest regions (such as a single U.S. state) are the same
  12. pretty print the output
@marycordova

View solution in original post

marycordova
SplunkTrust
SplunkTrust
| index=x OR index=y OR index=z src_ip!=192.168.* src_ip!=*:* user!=*test*
| lookup src_ips.csv src_ip as src_ip outputnew description
| where isnull(description)
| lookup users.csv attr1 as user outputnew email 
| lookup users.csv attr2 as user outputnew email <etc iterate as necessary or however you do user normalization>
| eval src_user='email'
| dedup src_user src_ip index _time
| rex field=_time "(?<dest_time>^\d+)"
| rename src_ip as dest_ip
| sort 0 src_user dest_time
| streamstats values(dest_ip) as src_ip values(dest_time) as src_time by src_user window=1 current=false <the window and current options are KEY here>
| where isnotnull(src_ip) AND (NOT 'src_ip'=='dest_ip')
| `geodistance(src_ip,dest_ip)`
| where mph>575 AND 'src_region'!='dest_region'
| stats values(src_ip) as src_ip values(dest_ip) as dest_ip list(src_city) as src_city list(src_region) as src_region list(src_country) as src_country list(dest_city) as dest_city list(dest_region) as dest_region list(dest_country) as dest_country avg(miles) as miles avg(hours) as hours avg(mph) as mph by src_user
| eval miles=round(miles,2)
| eval km=round(km,2)
| eval hours=round(hours,2)
| eval mph=round(mph,2)
| eval kph=round(kph,2)
| eval src_ip=mvdedup(split((mvjoin((mvzip(src_ip,dest_ip)),",")),","))
| eval src_city=mvdedup(split((mvjoin((mvzip(src_city,dest_city)),",")),","))
| eval src_region=mvdedup(split((mvjoin((mvzip(src_region,dest_region)),",")),","))
| eval src_country=mvdedup(split((mvjoin((mvzip(src_country,dest_country)),",")),","))
| table src_user src_ip src_city src_region src_country miles km hours mph kph
@marycordova
0 Karma

marycordova
SplunkTrust
SplunkTrust
  1. base search: as many indexes/data sources as might contain authentication, authorization, or access data that may be relevant for monitoring potential unauthorized or malicious activity
  2. filter out garbage IP addresses: src_ip!=<rfc 1918 addresses> src_ip!=<whatever other garbage you don't need> maybe you don't really want to try to deal with IPv6 addresses, maybe you have a lookup that you can drop all your known public ip space
  3. filter out garbage user data: user!=*test* user!=<whatever other garbage you don't need>
  4. normalize user: in my case I take whatever user attribute is in the base search, username or email address or whatever, and then query my IAM provider to return a standard value type for my user field, for example email address
  5. dedup for each user: from your base search you want each unique ip address and _time for each individual user, for me this took about 250K events down to about 30K, you would think including _time in the dedup would render the dedup useless but in this case it did not and _time is a critical field, you may also want to include index or sourcetype (whichever is applicable)
  6. prep for streamstats, the "current" data becomes the "destination": dest_time, dest_ip
  7. prep for streamstats, sort your data for each user and _time: streamstats needs to work a single user's data _time sorted so the results aren't broken into different sets (default limit of 10K)
  8. streamstats to create the "source" for each "destination": from the users previous event src_time, src_ip are generated (default limit of 10K)
  9. filter out garbage streamstats data: the first event for every user will have a null src_ip, and events with the same src and dest ip addresses are useless
  10. calculate geo statistics: the geodistance macro will take src_ip and dest_ip as arguments and return the city, region and country for each ip, the distance traveled in km and miles, hours it took to travel, and the kph and mph
  11. additional filtering logic: in my case I dropped all events where mph are less than the speed of a commercial airplane as well as any events where the src and dest regions (such as a single U.S. state) are the same
  12. pretty print the output
@marycordova

marycordova
SplunkTrust
SplunkTrust

create a new macro with no arguments: "geodistance"

alt text

iplocation src_ip 
| rename lat as src_lat lon as src_lon City as src_city Region as src_region Country as src_country 
| iplocation dest_ip 
| rename lat as dest_lat lon as dest_lon City as dest_city Region as dest_region Country as dest_country 
| search src_lat!="," src_lon!="," dest_lat!="," dest_lon!="," 
| eval lat=(dest_lat-src_lat)*pi()/180 
| eval lon=(dest_lon-src_lon)*pi()/180 
| eval dest_rad=dest_lat*pi()/180 
| eval src_rad=src_lat*pi()/180 
| eval a=pow(sin(lat/2),2)+pow(sin(lon/2),2)*cos(dest_rad)*cos(src_rad) 
| eval miles=(12742*atan2(sqrt(a),sqrt(1-a)))*0.621371 
| eval km=12742*atan2(sqrt(a),sqrt(1-a)) 
| eval hours=if(((dest_time-src_time)/60/60)=0,".01",(abs(((dest_time-src_time)/60/60)))) 
| eval mph=miles/hours 
| eval kph=km/hours 
| fillnull value=""
@marycordova
0 Karma

MuS
Legend

As an add-on here is my macro I use to get distances based on lat's and lon's 🙂

[distance(4)]
args = lat1,lon1,lat2,lon2
definition = eval rlat1 = pi()*$lat1$/180, rlat2=pi()*$lat2$/180, rlat = pi()*($lat2$-$lat1$)/180, rlon= pi()*($lon2$-$lon1$)/180\
| eval a = sin(rlat/2) * sin(rlat/2) + cos(rlat1) * cos(rlat2) * sin(rlon/2) * sin(rlon/2) \
| eval c = 2 * atan2(sqrt(a), sqrt(1-a)) \
| eval distance = 6371 * c |  fields - a c r*
iseval = 0

It is part of my Dark Sky app featured in this .conf2017 talk https://conf.splunk.com/files/2017/slides/take-a-talk-into-the-art-of-dark-sky-photography-with-a-sp...

cheers, MuS

wrangler2x
Motivator

Interesting.

0 Karma

marycordova
SplunkTrust
SplunkTrust

thanks! 🙂

@marycordova
0 Karma
Get Updates on the Splunk Community!

Take Your Breath Away with Splunk Risk-Based Alerting (RBA)

WATCH NOW!The Splunk Guide to Risk-Based Alerting is here to empower your SOC like never before. Join Haylee ...

SignalFlow: What? Why? How?

What is SignalFlow? Splunk Observability Cloud’s analytics engine, SignalFlow, opens up a world of in-depth ...

Federated Search for Amazon S3 | Key Use Cases to Streamline Compliance Workflows

Modern business operations are supported by data compliance. As regulations evolve, organizations must ...