Dynamic-link library (DLL)
when Windows Side-by-Side (WinSxS)
not explicit about the characteristics of DLLs being loaded by a
program. In layman’s terms, DLL side-loading can allow an attacker to
trick a program into loading a malicious DLL. If you are interested in
learning more about how DLL side-loading works and how we see
attackers using this technique, read through our whitepaper.
hijacking occurs when an attacker is able to take advantage of the
Windows search and load order, allowing the execution of a malicious
DLL, rather than the legitimate DLL.
DLL side-loading and hijacking has been around for years; in fact,
FireEye Mandiant was one of the
to discover the DLL side-loading technique along with
search order hijacking back in 2010. So why are we still writing
a blog about it? Because it’s still a method that works and is used in
real world intrusions! FireEye Mandiant still identifies and observes
threat groups using DLL abuse techniques during incident response (IR)
engagements. There are still plenty of signed executables vulnerable
to this, and our red team has weaponized DLL abuse techniques to be
part of our methodology. For detection and preventative measures on
DLL abuse techniques, see the “Detection and Preventative Measures”
section in this blog post.
Even though DLL abuse techniques are not new or cutting edge, this
blog post will showcase how the
Mandiant red team uses
Intelligence to expedite the research phase of identifying
vulnerable executables, at scale! We will also walk you through how to
discover new executables susceptible to DLL abuse and how the FireEye
Mandiant red team has weaponized these DLL abuse techniques in its
The DueDLLigence tool was initially released to be a
for application whitelisting bypasses, but given the nature of
unmanaged exports it can be used for DLL abuse techniques as well.
A benefit of being part of the red team at FireEye Mandiant is
having access to a tremendous amount of threat intelligence; Our
organization’s incident response and intelligence consultants have
observed, documented, and analysed the actions of attackers across
almost every major breach over the past decade. For this project, the
FireEye Mandiant red team asked the FireEye Technical Operations and
Reverse Engineering Advanced Practices (TORE AP) team to leverage
FireEye Intelligence and provide us with all DLL abuse techniques used
by attackers that matched the following criteria:
Once the results were provided to the red team, we started
weaponizing the intelligence by taking the approach outlined in the
rest of the post, which includes:
In many cases it is possible to execute code within the context of a
legitimate
Executable (PE) by taking advantage of insecure library
references. If a developer allows
dynamically then that PE will also look in the current directory for
the library DLL. This behavior can be used for malicious purposes by
copying a legitimate PE to a directory where the attacker has write
access. If the attacker creates a custom payload DLL, then the
application will load that DLL and execute the attacker’s code. This
can be beneficial for a red team: the PE may be signed and have the
appearance of trust to the endpoint security solution (AV/EDR), it may
bypass application white listing (AWL) and can confuse/delay an
investigation process.
In this section we will look at one example where we identify the
conditions for hijacking a PE and implement the requirements in our
payload DLL. For this test case we will use a signed binary
PotPlayerMini (MD5:
chosen since it has been used by attackers dating back to 2016.
It is possible to determine which libraries and exports a PE
requires through static analysis with tools such as
Figure 1, for example, shows that PotPlayerMini tries to load a DLL
called “PotPlayer.dll”.
Figure 1: Static Analysis of DLL's loaded
by PotPlayerMini
Where static analysis is not feasible or desirable it may be
possible to use a hooking framework such as
In Figure 2 we used API Monitor to see this same DLL loading
behavior. As you can see, PotPlayerMini is looking for the
this point, we have validated that PotPlayerMini is susceptible to DLL
search order hijacking.
Figure 2: Dynamic Analysis of DLL's
loaded by PotPlayerMini
After identifying potentially vulnerable library modules we need to
apply a similar methodology to identify which exports are required
from the module PE. Figure 3 shows a decompiled view from
PotPlayerMini highlighting which exports it is looking for within the
GetProcAddress functions using static
analysis. Figure 4 shows performing this same analysis of exports in
the PotPlayerMini application, but using dynamic analysis instead.
Figure 3: Static Analysis of exports in
PotPlayerMini DLL
Figure 4: Dynamic Analysis of exports in
PotPlayerMini DLL
In our case the payload is a .NET DLL which uses
to satisfy all export requirements from the binary as shown in Figure
5. This is because the .NET UnmanagedExports library does not support
not exported. All export requirements need to be satisfied to ensure
the DLL has all the functions exported which the program accesses via
GetProcAddress or import address table
(IAT). These export methods will match those that were observed in the
static and dynamic analysis. This may require some trial and error
depending on the validation that is present in the binary.
Figure 5: Adding export requirements in
.NET DLL
Once we execute the binary, we can see that it successfully executes
our function as shown in Figure 6.
Figure 6: Executing binary susceptible to
DLL abuse
When writing a payload DLL in C/C++ it is possible to hijack control
flow in
to enumerate and satisfy all needed exports as previously described.
There also may be cases where the DLL does not have any exports and
can only be hijacked via the DllMain entry point.
An example of this can be shown with the Windows Media Player Folder
Sharing executable called wmpshare.exe. You
can copy the executable to a directory out of its original location
(C:\Program Files (x86)\Windows Media
Player) and perform dynamic analysis using API Monitor. In
Figure 7, you can see that the wmpshare.exe program uses the
path to the DLL. When this happens, the
directory in which the process was created (present working
directory). Full details on the search order used can be found in the
Figure 7: Viewing LoadLibrary calls in wmpshare.exe
Since it does not specify an explicit path, you can test if it can
be susceptible to DLL hijacking by creating a blank file named “
the wmpshare.exe file. Now when running the
wmpshare executable in API Monitor, you can see it is first checking
in its current directory for the wmp.dll
file, shown in Figure 8. Therefore, it is possible to use this binary
for DLL hijacking.
Figure 8: Viewing LoadLibrary calls in
wmpshare.exe with dummy dll present
Figure 9 shows using the wmpshare executable in a weaponized manner
to take advantage of the DllMain entry point
with a DLL created in C++.
Figure 9: Using the DllMain entry point
In addition to weaponizing the FireEye intelligence of the
executables used for DLL abuse by attackers, the FireEye Mandiant red
team performed research to discover new executables susceptible to
abuse by targeting Windows system utilities and third-party applications.
The FireEye Mandiant red team used the methodology previously
described in the Collecting and Weaponizing FireEye Intelligence
section to look at Windows system utilities present in the
susceptible to DLL abuse techniques. One of the system utilities found
was the deployment image servicing and management (DISM) utility
(Dism.exe). When performing dynamic
analysis of this system utility, it was observed that it was
attempting to load the DismCore.dll file in
the current directory as shown in Figure 10.
Figure 10: Performing dynamic analysis of
Dism utility
Next, we loaded the DISM system utility into API Monitor from its
normal path (C:\Windows\System32) in order
to see the required exports as shown in Figure 11.
Figure 11: Required exports for DismCore.dll
The code shown in Figure 12 was added to
validate that the DLL was vulnerable and could be ran successfully
using the DISM system utility.
Figure 12: Dism export method added to DueDLLigence
The FireEye Mandiant red team also targeted executable files
associated with common third-party applications that could be
susceptible to DLL abuse. One of the executable files discovered was a
Tortoise SVN utility (
of this Tortoise SVN utility, it was observed that it was attempting
to load crshhndl.dll in the current
directory. The export methods are shown in Figure 13.
Figure 13: Performing dynamic analysis of SubWCRev.exe
The code shown in Figure 14 was added to
validate that the DLL was vulnerable and could be ran successfully
using the Tortoise SVN utility.
Figure 14: SubWCRev.exe export methods
added to DueDLLigence
Having a standalone trusted executable allows the red team to simply
copy the trusted executable and malicious DLL to a victim machine and
bypass various host-based security controls, including application
whitelisting. Once the trusted executable (vulnerable to DLL abuse)
and malicious DLL are both in the same present working directory, the
executable will call the corresponding DLL within the same directory.
This method can be used in multiple phases of the attack lifecycle as
payload implants, including phases such as establishing persistence
and performing lateral movement.
In this example, we will be using the Windows system utility
Utilities section as our executable, along with a DLL generated by
conjunction with
establish persistence on a target system. First, the DISM system
utility and malicious DLL are uploaded to the target system as shown
in Figure 15.
Figure 15: Uploading payload files
Then we use
startup folder
persistence, which uses our DISM system utility and associated DLL
as shown in Figure 16.
Figure 16: Adding startup folder
persistence with SharPersist
After the target machine has been rebooted and the targeted user has
logged on, Figure 17 shows our Cobalt Strike C2 server receiving a
beacon callback from our startup folder persistence where we are
living in the Dism.exe process.
Figure 17: Successful persistence callback
We will continue using the same DISM system utility and DLL file for
lateral movement. The HOGWARTS\adumbledore
user has administrative access to the remote host
DISM system utility and the associated DLL file via the SMB protocol
to the remote host as shown in Figure 18.
Figure 18: Transferring payload files to
remote host via SMB
Then we setup a SOCKS proxy in our initial beacon, and use
Impacket’s
execute our payload via the Windows Management Instrumentation (WMI)
protocol, as shown in Figure 19 and Figure 20.
proxychains python wmiexec.py |
Figure 19: Executing payload via WMI with
Impacket’s wmiexec.py
Figure 20: Output of executing command
shown in Figure 19
We receive a beacon from the remote host, shown in Figure 21, after
executing the DISM system utility via WMI.
Figure 21: Obtaining beacon on remote host
Detailed prevention and detection methods for DLL side-loading are
well documented in the
mentioned in the DLL Abuse Techniques Overview. The whitepaper breaks
it down into preventative measures at the software development level
and goes into recommendations for the endpoint user level. A few
detection methods that are not mentioned in the whitepaper include:
These detection methods are difficult to implement at scale, but
possible to utilize. That is exactly why this old technique is still
valid and used by modern red teams and threat groups. The real problem
that allows this vulnerability to continue to exist has to do with
software publishers. Software publishers need to be aware of DLL abuse
techniques and know how to prevent such vulnerabilities from being
developed into products (e.g. by implementing the mitigations
discussed in our
Applying these recommendations will reduce the DLL abuse opportunities
attackers use to bypass several modern-day detection techniques.
Microsoft has provided some great resources on
security and
a DLL hijacking vulnerability.
Threat intelligence provides immense value to red teamers who are
looking to perform offensive research and development and emulate
real-life attackers. By looking at what real attackers are doing, a
red teamer can glean inspiration for future tooling or TTPs.
DLL abuse techniques can be helpful from an evasion standpoint in
multiple phases of the attack lifecycle, such as persistence and
lateral movement. There will continue to be more executables
discovered that are susceptible to DLL abuse and used by security
professionals and adversaries alike.