Mr. Robot, Mimikatz and Lateral Movement

In Mr. Robot‘s episode 9 of season 2 (13:53), Angela Moss needs to obtain the Windows domain password of her superior, Joseph Green, in order to download sensitive documents that would potentially incriminate EvilCorp. Since her attack requires physical access to his computer, she starts with a good old-fashioned social engineering attack to get the only currently present employee in the office to leave.

Angela “obtaining” logon credentials

Once in his office, she uses a USB Rubber Ducky, a fast and automated keyboard emulator, to obtain Joseph’s clear text password using mimikatz. Please note that there are some holes in this scene which I will get into later. For now we’ll assume that she was able to obtain his credentials by having physical access to his computer.

USB Rubber Ducky
USB Rubber Ducky

After she gets back to her workstation, she analyzes the capture which reveals Joseph’s password: holidayarmadillo. Not the best password, but for this particular attack the quality of the password wouldn’t have mattered anyways. Mimikatz was (is) able to get the password from memory without utilizing brute force or dictionary techniques. Once she has the password, she logs off and logs back on with Joseph’s user and downloads the documents she needs.

Mimikatz Output
Mimikatz Output

As somebody who helps our users improve the security of their networks, I of course immediately contemplated how this attack could have been detected with EventSentry. Since most users only log on to one workstation on any given day with their user account, Angela logging in with Joseph’s account (resulting in “joseph.green” logging on to two different workstations) would actually be an easy thing to detect.

Lateral Network Movement
Lateral Network Movement

Introduced with v3.4, collector-side thresholds allow the real-time detection of pretty much any user activity that originates from an event, for example user logons or process launches. You can tell EventSentry that (physical) logons for any user on more than one host (within a given time period – say 9 hours) should trigger an alert (aka as “lateral movement”). Had this been in place at EvilCorp, Angela logging in as Joseph would have immediately triggered an alert. With the right procedures in place, countermeasures could have been taken. Of course most viewers wouldn’t want Angela to be caught, so please consider my analysis strictly technical. Watch a short video on lateral detection with EventSentry here.

So what’s the hole? Well, the rubber ducky (mimikatz, really) requires access to an active logon session, which Angela most likely didn’t have. It looked like Joseph had been out of the office for a while, so his computer was likely either locked or turned off, rendering any attack based on mimikatz useless. Mimikatz – since it obtains passwords from memory – only works if the computer is unlocked. And had the computer been unlocked then she could have just downloaded the files from his computer – although this would have been even more risky with people walking around the office.

Cyber attacks are becoming more potent every year and are often sponsored by powerful criminal gangs and/or governments. It’s important that companies employ multiple layers of defense to protect themselves (and their customers) from these increasingly sophisticated and destructive attacks.

EventSentry is the only monitoring solution that utilizes robust agent-based technology that goes beyond logs, enabling the fusion of real-time log monitoring with in-depth system monitoring to not only detect but also react to attacks and anomalies. See for yourself and download a free evaluation of EventSentry.

Securing Exchange Server OWA & ActiveSync – Proactive Security with EventSentry

Almost every company which runs Microsoft Exchange Server needs to make port 443 available to the Internet in order to provide their users access to email via their mobile devices or OWA.

Since both OWA & ActiveSync utilize Active Directory for authentication, exposing OWA/ActiveSync to the Internet indirectly exposes Active Directory as well. While user lockout policies provide some protection against brute force attacks, additional protection methods should be employed. Furthermore, password spraying attacks may be use to circumvent lockout policies – something that would be more likely to succeed in larger organizations.

With the proper auditing enabled (Logon/Logoff – Logon (Failure)) and EventSentry installed however, we can permanently block remote users / hosts who attempt to log on too many times with a wrong password. Setting this up is surprisingly simple:

  1. Windows: Enable (or verify) Auditing
  2. EventSentry: Setup action which creates firewall block rule
  3. EventSentry: Setup filter looking for 4625 Audit Failure events

Bonus: This procedure works with the free version of EventSentry (EventSentry Light) and can be applied to any IIS-based web site which uses authentication.

Windows Auditing

In the group policy settings that affect the server running OWA, make sure that auditing for Failure events in the Audit Logon sub category of the Logon/Logoff category is enabled (of course you can audit success events as well). If you are running the full version of EventSentry v3.4 or later then you can verify all effective audit settings on the Audit Policy Status page for example.

Enabling correct auditing for logon failures

Creating an Action

Since event 4625 contains the IP address of the remote host, the easiest way to subsequently block it is to run the netsh command. In the management console, create a new action by clicking on the “Action” header in the ribbon and selecting the process action as its type. See the screenshot below:

Triggering netsh to block an IP address

The following command line will work in EventSentry v3.4 and later:

advfirewall firewall add rule name="$STRIpAddress $YEAR-$MONTH-$DAY -- automatic block by EventSentry" dir=in interface=any action=block remoteip=$STRIpAddress/32

If you are running EventSentry v3.3 or earlier then you will need to use the $STR20 variable instead:

advfirewall firewall add rule name="$STR20 $YEAR-$MONTH-$DAY -- automatic block by EventSentry" dir=in interface=any action=block remoteip=$STR20/32

The difference here is that v3.4 and later can refer to insertion string variables by name, making the action more universal and potentially applicable to any event that uses the same field name.

When this action is triggered, it will extract the IP address from the event and block it from the system entirely.

Creating a Filter

Create an event log filter which matches Audit Failure events from the Security event log with event id 4625, where insertion string 19 matches the w3wp.exe process (C:\Windows\System32\inetsrv\w3wp.exe). This ensures that only users accessing the host via the web will be subject to blocking. The screenshot below shows the configuration:

Adding a firewall rule from a 4625 audit failure event

This filter can either be added to an existing package or added to a new package that is assigned only to the Exchange server. If the filter is added to an existing package that applies to servers other than the Exchange server, then the computer field of the filter can be used to ensure the filter is evaluated only on the desired host. Select the action created in the previous step.

Since users may occasionally enter an incorrect password I recommend setting up a threshold so that remote IPs are only blocked after 3 or more failed logon attempts. Threshold are configured by clicking on the “Threshold” tab (see the blue “i” above) and an example configuration is shown below. Feel free to adjust the threshold to match your users ability to enter their password correctly :-). Insertion string 20 – which represents the IP address of this event – was selected in the threshold matching section to ensure that each IP address has its own, unique threshold. Note: The event logging settings shown are optional.

Trigger process after 3 failed logon attempts.

Save/deploy or push the configuration to the mail server.

 

Considerations

Triggering a system process from external input is something we should always do with caution. For example, if Windows has an upper limit to the maximum number of rules that can be added, then an attacker could launch a DoS attack IF they had the ability to launch attacks from different IP addresses. Launching a DoS attack from the same IP won’t be possible once they are blocked. You can mitigate this risk by applying a threshold to the EventSentry action calling netsh.exe, for example by limiting it to 100 / hour. This would still provide sufficient protection while also ensuring that only 100 rules could be added per hour (thresholds can be set by clicking on the “Options” button on an action). A regular audit of the netsh execution (e.g. via Process Tracking) would quickly show any sort of abuse.

Over time the number of firewall rules added to the mail server could become rather large, which is why the rules are created with a date appended. This makes managing these rules easier, and the name can also be adapted in the action by changing the “rule name” parameter. The screenshot below shows the inbound firewall rules after two IPs have been blocked:

List of inbound firewall rules

If manual cleanup of firewall rules is not desirable or an option, then the netsh command can also be wrapped into a script which would erase the firewall rule again after a timeout (e.g. 15 minutes). The script could look like this:

"C:\windows\system32\netsh.exe" advfirewall firewall add rule name="%1 %2 -- automatic block by EventSentry" dir=in interface=any action=block remoteip=%1/32
timeout /t 900
"C:\Windows\system32\netsh.exe" advfirewall firewall delete rule name="%1 %2 -- automatic block by EventSentry"

In this case you would call the wrapper script instead of the netsh.exe process directly (General Options – Filename) and use the string below as the arguments:

$STRIpAddress $YEAR-$MONTH-$DAY

To keep things simple you can just make the script an embedded script (Tools menu) and reference the script. The timeout value (120 in the above example) is the duration seconds the remote IP will be be blocked. If you want to block the IP for an hour then you would set the timeout value to 3600 instead. When going this route I strongly suggest clearing both event log check boxes in the Options dialog of the action.

Happy Blocking!