Detecting CONTI CobaltStrike Lateral Movement Techniques - Part 2
Detection opportunities on lateral movement techniques used by CONTI ransomware group using CobaltStrike.
Last updated
Detection opportunities on lateral movement techniques used by CONTI ransomware group using CobaltStrike.
Last updated
In this second and last part of detecting CONTI lateral movement techniques I will go through the rest of CobaltStrike's built-in capabilities documented in the CONTI leak.
In the first blog post I tried to cover the jump command capabilities and detection opportunities where we compared them to some built-in windows utilities.
For the first part, please visit : Detecting CONTI CobaltStrike Lateral Movement Techniques - Part 1
WMI is Microsoft's implementation of Web-Based Enterprise Management (WBEM) which is an industry initiative to develop a standard technology for accessing management information in an enterprise environment and CIM (Common Information Model) which is an open standard from the Distributed Management Task Force (DMTF). CIM provides a common definition of management information for systems, networks, applications, and services.
WMI can be used over RPC/WinRM protocol or RPC/DCOM. In this introduction I will be focusing on RPC/DCOM.
Data in WMI is grouped into WMI classes. WMI classes are then grouped into WMI namespaces. Most of the WMI classes exist under the root\cimv2 WMI namespace.
In summary each Namespace contains Classes which have:
Methods : Actions that can be taken.
Properties : Information that can be retrieved.
Instances : Instances of the class objects (services, Processes, Disks) each instance with Methods and Properties.
Events : Actions that WMI can monitor for and take action when they happen.
WMI leverages DCOM server and client interfaces to communicate over the network between Windows Management Instrumentation Remote Protocol clients and servers.
When it comes to lateral movement one of my favorite data sources to check first is Zeek. Upon running the simulated lateral movement attack using CobaltStrike built-in command remote-exec wmi, the following telemetry was generated by Zeek.
zeek.dce_rpc.endpoint column values are the interfaces while zeek.dce_rpc.operation are the methods defined in WMI and DCOM documentations. This is very helpful in order to understand how WMI looks like from a network perspective. Zeek can identify these GUIDs related to IWbem interfaces. A full list is documented in GitHub source code here.
IObjectExporter::ServerAlive : First we can see RPC binding information calls to the IObjectExporter interface using methods ServerAlive or ServerAlive2 to determine server aliveness. Deciding the method is related to the COMVERSION in use.
IRemoteSCMActivator::RemoteCreateInstance : The DCOM client MUST support the Activation and OXID Resolution DCOM mechanisms for creating and resolving object references. Activation mechanism can be achieved through two interfaces and three different methods, IActivation::RemoteActivation , IRemoteSCMActivator::RemoteCreateInstance , or IRemoteSCMActivator::RemoteGetClassObject.
IRemUnknown2::RemQueryInterface : Every object can be bound to one or multiple interfaces. An Object reference counter is used to keep track of a Component Object Model (COM) objects. For acquiring additional interfaces on the object IRemUnknown::RemQueryInterface and IRemUnknown2::RemQueryInterface calls are used.
An object reference is represented on the wire by a marshaled form called OBJREF.
IWbemLevel1Login::NTLMLogin : According to MS-WMI documentation, during protocol initialization, The client MUST call the IWbemLevel1Login::NTLMLogin method.
IWbemServices::ExecMethod : This call will return an interface pointer to IWbemServices management services where methods like GetObject which retrieves a CIM class or a CIM instance and ExecMethod which executes a CIM method that is implemented by a CIM class or a CIM instance, can be used.
IRemUnknown2::RemRelease : The release sequence is then called to decrement the reference counter
Bellow is a mind-map where I tried to summarize the different interfaces and method used during WMI remote calls. This will help understand the telemetry recorded by Zeek in order to identify the best calls to focus our detections on.
As stated in the MS-WMI documentation, during protocol initialization, the client MUST call the IWbemLevel1Login::NTLMLogin method. This is a good indication of WMI usage over the network. However, a good baseline of users and assets with authorization to use WMI accompanied with a well defined change management process will significantly improve your detection success rate. IWbemServices::ExecMethod and IWbemServices::GetObject calls are also good indications of WMI accessing web-based management services.
Zeek Telemetry:
| Log File | Endpoint | Operation |
|---|---|---|
DCE-RPC | IWbemLevel1Login |
|
DCE-RPC | IWbemServices |
|
DCE-RPC | IWbemServices |
|
CobaltStrike has a built-in lateral movement module called remote-exec which supports three commands : wmi, winrm, and psexec. Remote-Exec module is used to execute a command on a host remotely and doesn't pop a beacon unless it is used for that particular purpose by first uploading a script or a beacon file then execute it via remote-exec commands and use link or connect commands to assume control of the target.
In this section I will be exploring some generated telemetries from the endpoint perspective using wmi command.
wmiprvse.exe process is spawned with the command line C:\\Windows\\system32\\wbem\\wmiprvse.exe -secured -Embedding and parent command line C:\\Windows\\system32\\svchost.exe -k DcomLaunch.
EID 5857 was generated to report the start of WMI provider cimwin32.dll. There are several WMI providers. This is not very useful because WMI usage can be verbose.
The command is executed within the context of *WmiPrvSE.exe*.
By default, WMI uses a randomly selected dynamic port range for TCP between 49152 and 65535.
| EID | Action | Provider | Comment |
|---|---|---|---|
1 | Process Creation | Microsoft-Windows-Sysmon |
|
1 | Process Creation | Microsoft-Windows-Sysmon |
|
3 | Network Connection | Microsoft-Windows-Sysmon |
|
Detecting malicious usage of WMI relies heavily on WmiPrvse.exe abnormal child processes behavior. However, some approaches can be taken to improve your detections. For example if you have a SCCM server, you might consider whitelisting the following paths in your process command arguments (Reference):
Keep in mind that attackers might still abuse these paths to evade detections so baselining your assets, source IPs and users that are allowed to use WMI remotely is recommended to increase detection resilience.
By default only Local Administrators or Domain Admins can read WMI class information so in order to further refine your access control policies you can limit regular users permissions by adding them to the Distributed COM Users group and the Performance Monitor Users group.
In the leaked CONTI documentation, we noticed a lot of wmic.exe usage for remote command execution across multiple assets. For example, they use a batch file called WMI.BAT with the following command to spread a binary file across multiple hosts.
Or interact with beacon through shell command to dump credentials :
WMIC.EXE is one of the Windows built-in utilities that leverages WMI protocol for command execution. For detection opportunities we can look for :
From source point of view, process command line arguments process call create and for WMI remoting we can add the argument /node. You can use @wietze Windows Command Line Obfuscation project to validate command line variations for more resilient detections based on process creation events.
EID 4648 A logon was attempted using explicit credentials where the process name is svchost.EXE and service class RPCSS*. This event is a good DFIR artifact for differentiating between the original account and the account specified in the wmic command (In my case I didn't specify any credentials).
A service principal name (SPN) is the name by which a Kerberos client uniquely identifies an instance of a service for a given Kerberos target computer. There are multiple SPN registrations :
HTTP/hostname.contoso.com like when using PowerShell Remoting via Enter-PSSession
WSMAN/hostname.contoso.com like when using WinRM for Remoting
CIFS/hostname.contoso.com like when using PsExec
HOST/hostname.contoso.com for any service running on the computer with hostname HOSTNAME
The RPCSS service is the Service Control Manager for COM and DCOM servers. It performs object activations requests, object exporter resolutions and distributed garbage collection for COM and DCOM servers (source). HOST service can also be used for remotely executing commands on the target system via WMI (source).
On the destination, as previously explained, looking for abnormal behavior of WmiPvSE.exe like spawning PowerShell.exe and Cmd.exe with suspicious arguments would be effective. (see previous table Endpoint for more details)
The table bellow displays WMIC related telemetry generated from the source host :
| EID | Action | Provider | Comment |
|---|---|---|---|
1 | Process Creation | Microsoft-Windows-Sysmon |
|
3 | Network Connection | Microsoft-Windows-Sysmon |
|
4648 | Authentication | Microsoft-Windows-Security-Auditing |
|
4648 | Authentication | Microsoft-Windows-Security-Auditing |
|
The following rules present some ideas about detecting malicious WMI behavior.
Atomic Red Team provides a good resource to test your WMI detections
EDR Testing Script :
Test the accuracy of Endpoint Detection and Response (EDR) software with simple script which executes various ATT&CK/LOLBAS/Invoke-CradleCrafter/Invoke-DOSfuscation payloads
To provide more details about the WMI activity for your DFIR engagements, you can use ETW. To enable the event tracing of WMI, you can use the command line:
Be aware that ETW was made for debugging and enabling WMI event tracing features might generate a lot of data which will be stopped after reaching a certain size/duration limit.
remote-exec winrm command is similar to jump winrm64 in command execution under the context of wmsprovhost.exe except that it was not made for creating and maintaining a remote session hence wsmprovhost.exe terminates after execution.
Generated telemetry on the destination :
| EID | Action | Provider | Comment |
|---|---|---|---|
1 | WSMan Session Creation | Microsoft-Windows-Sysmon |
|
3 | WSMan Session Creation | Microsoft-Windows-Sysmon |
|
17 | Pipe Created | Microsoft-Windows-Sysmon |
|
4656 | Process Access | Microsoft-Windows-Security-Auditing |
|
400 | PowerShell Session Start | PowerShell |
|
91 | WSMan Session Creation | Microsoft-Windows-WinRM | |
31 | WSMan Session Creation | Microsoft-Windows-WinRM | WSMan Session Created Successfully |
142 | WSMan Operation Failure | Microsoft-Windows-WinRM | Helpful when WinRM is not enabled on the targeted host |
remote-exec psexec command creates and start a service remotely with random Service Name and the passed on command as Service File Name. The main difference between this feature and jump psexec or jump psexec64 is that remote-exec psexec does not generate a service executable and upload it to the target. As noticed before, CobaltStrike's service file spawns rundll32.exe with no arguments which is suspicious.
Monitoring services.exe child process for malicious behavior like spawning system shells cmd.exe and powershell.exe or other discovery binaries like whoami.exe, systeminfo.exe, net.exe,...etc would be effective against this type of attack.
In the CONTI leaked documentation, the playbook shows the usage of this module to dump lsass.exe memory via comsvcs.dll
This detection rule from Elastic should be enough to detect such behavior.
See previous blog for more details on CobaltStrike psexec built-in capabilities detection.
As defined by MITRE in ATT&CK framework:
Adversaries may "pass the hash" using stolen password hashes to move laterally within an environment, bypassing normal system access controls. Pass the hash (PtH) is a method of authenticating as a user without having access to the user's cleartext password. This method bypasses standard authentication steps that require a cleartext password, moving directly into the portion of the authentication that uses the password hash.
CobaltStrike has a built-in module called pth to perform pass-the-hash attack using Mimikatz's sekurlsa:pth module. As stated by CobaltStrike creator himself this is not OpSec safe since it presents low hanging detection opportunities for defenders.
PTH module has a hardcoded command that contains suspicious sequence of arguments such as *cmd.exe /c echo > \\.\\pipe\* . Monitoring process creation events with such arguments would be effective against CobaltStrike's way of implementing and automating pass-the-hash attack. Keep in mind attackers can always use Mimikatz PTH module where they can change these properties.
Another key event for detecting pass the hash is EID 4624 with logon type 9 (NewCredentials), logon process seclogo and Authentication Package Negotiate.
PTH detection observations :
| EID | Action | Provider | Comment |
|---|---|---|---|
1 | Process Creation | Microsoft-Windows-Sysmon |
|
4624 | Authentication | Microsoft-Windows-Security-Auditing |
|
The CONTI leaked documentation shows RDP being used several time for manual access whether to dump lsass process memory using task manager or export credentials from users profiles and keyloggers data. This is not an exploitation of the RDP service itself since the attacker already got their hands on user's credentials, so in this case maintaining a good RDP users policy will help creating a baseline and detecting related violations. EID 4825 A user was denied the access to Remote Desktop can be helpful in this matter.
I previously created this mind map for RDP DFIR Authentication event logs that can be observed in your environment when using RDP with and without NLA enabled.
The mind map was pushed to a great GitHub project started by Andrew Rathbun (@bunsofwrath12) here. The RDP mind map can be found following this link :