Monthly Archives: June 2013

Using Splunk to watch for new binaries

The method presented below can be used to track any log attribute in Splunk; this example demonstrates watching MD5 hashes of executed files and loaded modules.

I’ve enabled Process Auditing via the Group Policy Editor and configured WLS to provide MD5 hashes.

standardhashes

I also enabled the “ModuleMonitor” in WLS which tracks loaded modules by process

modulemonitoron

and configured it to provide MD5 hashes for these as well.

modulehashes

Now that we are receiving hashes for all executed files and loaded modules, let’s start tracking them in Splunk.

First we’ll need to create a lookup table, there are a few ways to do this, a quick way is simply:

| outputlookup md5tracker.csv

This will create an empty csv file named “md5tracker.csv”.

emptycsv

Next, we need to search for and add the desired data to the csv file. I like to preserve some of the metadata that WLS reports with each record for later use – avoid re-searching, etc.

index=windows MD5=* | dedup MD5 | lookup md5tracker.csv MD5 as MD5 OUTPUT FirstSeen as LookupFirstSeen | where NOT LookupFirstSeen LIKE “%” | eval FirstSeen=_time  | table FirstSeen, MD5, BaseFileName, CompanyName, FileDescription, FileVersion, InternalName, Language, Signed, Length | inputlookup md5tracker.csv append=t | dedup MD5 | outputlookup md5tracker.csv

OK, let’s break this down:

Find the desired records: index=windows MD5=*

Remove duplicates: dedup MD5

Lookup the MD5s in our lookup table, returning the date first seen: lookup md5tracker.csv MD5 as MD5 OUTPUT FirstSeen as LookupFirstSeen

Remove records that already exist (field will be non-null): where NOT LookupFirstSeen LIKE “%”

Preserve the time stamp as desired output field: eval FirstSeen=_time

Format the desired fields into a table: table FirstSeen, MD5, BaseFileName, CompanyName, FileDescription, FileVersion, InternalName, Language, Signed, Length

Bring all the old data in and append it: inputlookup md5tracker.csv append=t

Remove duplicates (just in case): dedup MD5

Write out the new + old data: outputlookup md5tracker.csv

After the first run, you should have the results from your chosen time period now stored in md5tracker.csv

firstresults

You’ll want to save this search

save1

and schedule it to run every x minutes for the last x minutes; I schedule mine for every 15 minutes.

save2

Once this is complete you’ll now have a search that keeps your lookup table up-to-date. Now what?

What you do next depends on how closely you feel this needs monitored. I run a second search every x minutes that alerts on all new entries in the last x minutes (based on the FirstSeen) field.

| inputlookup md5tracker.csv | where now()-FirstSeen < 2200 | table FirstSeen, MD5, BaseFileName, CompanyName, FileDescription, FileVersion, InternalName, Language, Signed, Length

This simply take the entire table and selects all entries in the last 2200 seconds (2200 / 60 = 36.6 minutes) and formats the results into a table. I scheduled it to run every 35 minutes with some overlap time (hence 2200 instead of 2100).

save3

I also like to take an export of the hashes every so often and check them against Team Cymru’s malware hash registry – https://hash.cymru.com/

| inputlookup md5tracker.csv | table MD5

Export the results from Splunk, open the file in a spreadsheet, and copy/paste them into Team Cymru’s lookup for a quick analysis. An enterprising person might also create a custom Splunk command that uses their DNS lookup service (https://www.team-cymru.org/Services/MHR/#dns) and puts the results into the lookup table itself…

I currently have 23,537 executable hashes and 131,885 module (dll, etc) hashes, and see a few new ones at most search intervals during normal business hours. After the initial gathering, the periodic alerts are easy to quickly review, and you’ll know everything that is running on your Windows hosts.