Threat Analysis Unit

IcedID: Analysis and Detection

IcedID, also known as BokBot, was first documented in 2017. While the denomination IcedID used to be only about the final banking trojan payload, it now commonly refers to the full infection chain characteristic of this threat. IcedID stood under the radar for a couple of years, and made the news again in 2019 for using steganography to hide its payload. 

Ever since IcedID loaders evolved and went through multiple steganography techniques. Some of the versions are known as “Photoloader” or lately “Gziploader”. The functionalities of the underlying banking trojan themselves have seen little evolutions over the years. 

This blogpost will focus on 3 parts: 

  • Description of an IcedID infection chain 
  • Detailed overview of a dropper document 
  • VMware IcedID threat coverage 

IcedID infection chain

Figure 1 IcedID infection chain

The most commonly observed attack chain for IcedID can be split into three parts:

  • Initial delivery: Malicious office document
  • 1st stage: Packed DLL
  • 2nd stage: Packed DLL + encrypted payload

Initial delivery:

The entry point is an email with a malicious payload attached. Usually, the malicious payload is stored in a password-protected ZIP, with the password written in the body of the email. This is done as an attempt to bypass automated analysis which is commonly done by many email security products.

Once unlocked, the ZIP archive contains an office document with a malicious payload. Lately, researchers have reported on IcedID using Jscript dropper, (you can watch malware analysis here) but this is out of the scope of this blogpost.

In scenarios where malicious office documents were leveraged, we observed two types of payloads:

  • VBA macro code, described later in this blogpost
  • XL4

Ultimately, the sample will download the 1st stage DLL and execute it using rundll32.

Stage 1:

The first stage is a single DLL executed with the help of rundll32, and acts as a filter deciding whether the victim is worth compromising further or not. This stage has two core functionalities:

  • Fingerprinting the OS and sending the data to the CnC
  • Downloading the next payload from the CnC and executing that payload

Below is a list of some of the data that the first stage payload will gather:

  • How long has the computer been up
  • How many processes are running
  • OS version
  • CPUID queried data
  • RDTSC timing loop
  • Account information
  • Hardware info

This data is then transmitted to the CnC (Figure 2) which likely blacklisted some values in order to avoid uninteresting target or identified sandbox/analyst boxes. If nothing prevents it on the CnC side, the DLL then downloads a blob of data that will ultimately become the 2nd stage files. This data is not downloaded as plaintext but hidden via steganography techniques. At the time of the analysis the data was hidden behind a fake GZIP header, and previously we also observed fake JPG images.

Figure 2 System fingerprinting data hidden amongst cookie values

Stage 2:

The 2nd stage consists of two components:

  • A DLL: a loader
  • An encrypted data blob: the banking trojan

The DLL is executed using the rundll32 executable. The path to the encrypted data blob, often named “license.dat” is provided via the command line (Figure 3).

Figure 3 Stage 2 command line

Upon execution, the DLL will parse its command line to find the path to the data blob, decrypt it and load it into memory. The banking trojan is then live and running. A good description of the trojan functionalities can be read in this article.


Sample hash: 9a93fc9f3606055fad6f7a2a9b0a848555d9e8d29eb3e5419a6803c315e8cba4

This section is an in-depth analysis of a malicious word document found during our investigation.

As described earlier, a user would access this document after opening an encrypted zip attached to an email. Once the document is opened the user is greeted with a banner asking to “Enable Content” in order to execute macros (Figure 4). A careless user clicking that button would trigger the execution of the malicious VBA macro code.

Figure 4 Macro embedded document

Below (Figure 5) is an excerpt of the VBA macro code stored in the malicious document, extracted using oletools.

Figure 5 VBA macro code snippet

The VBA macro looks simple and not obfuscated. However, the malware authors have used interesting tricks and techniques to hide the malicious code. Indeed, dynamic execution of this code highlights that the macro drops an HTA file in the computer public folder. The HTA content is not stored in the VBA code itself but is hidden behind the image giving the instructions, out of sight from victims and analysts, as shown in Figure 6. In previous versions, actors used to hide code within a UserForm as labels. This is not a new technique and has been observed in other campaigns.

Figure 6 Hidden HTA code revealed as document text


Figure 7 Code snippet fetching the HTA content

Figure 7 above shows the functions dedicated to extract the HTA file content, by reading the “ActiveDocument.Range.text” attribute.

In a similar fashion, other data needed by the VBA macro is hidden in the document properties.

Figure 8 Data hidden in the document properties.

As shown above, in Figure 8, the path of the dropped HTA file is stored in the “Title” property, and a portion of the CLSID is hidden in the “Subject” property of the document. Below in Figure 9 is the code that fetches these properties as well as drops the HTA file content on disk.

Figure 9 Code snippet to fetch document properties, write hta file and execute hta using COM

To execute this HTA file, the attacker used an interesting COM technique which is also used for parent process ID (PPID) spoofing.

Indeed the CLSID C08AFD90-F2A1-11D1-8455-00A0C91F3880 used in the code refers to the ShellBrowserWindow object class and can be used to execute any process as if it was executed from explorer.exe. In this case, mshta.exe is executed by winword.exe, but the parent process of mshta.exe will be explorer.exe. Adversaries use this technique in an attempt to defeat threat hunting activities or to bypass security products. This PPID Spoofing method is already explained in our previous blog post here. This blog post also provides guidance to hunt such anomalies.

In its raw form, the HTA code is obfuscated as shown below Figure 10:

Figure 10 Obfuscated (top) and de-obfuscated (bottom) HTA payload

Desobfucating the HTA content reveals that its only goal is to download a file from a remote server, and to write it on disk in the public folder, under the name “structButtonMemory.jpg”. Despite the “.jpg” extension, this file is actually a DLL, and gets executed via rundll32.exe calling the “PluginInit” exported function as an entry point. This DLL is the 1st stage DLL payload, also referred to as Gziploader.

As mentioned in the introduction overview, this 1st stage DLL will download a file purporting to be a gzip compressed file. The first stage DLL will then extract two files from the fake gzip file: another DLL (the 2nd stage) and an encrypted data blob, which will be saved with a “.dat” extension. The second stage DLL file is a bot loader which is executed with rundll32.exe, with the file path to the .dat file being provided as a parameter. This process decodes the main bot in memory and performs additional malicious activities.

VMware detection coverage

VMware EDR Carbon Black will alert as soon as the dropper document gets executed thanks to multiple fine-tuned rules.

Anomaly in winword.exe process

Executing HTA payload via PPID spoofing 

Rundll32 payload execution

VMware NDR and sandboxing capabilities protect the network from any stage of the infection.

Initial document analysis


1st stage DLL analysis


2nd stage DLL and final payload analysis



Through this blogpost, we detailed the different steps commonly involved in an IcedID infection and did a deep dive into one of its initial infection document. IcedID is a three stages threat involving two DLL loaders executed via rundll32.exe. While the last stage of the IcedID chain is a banking trojan, IcedID can also be used as a pivot point to deliver other threats such as Ransomware or even serve as a foothold for further lateral propagation.

Leveraging both the VMware EDR and NDR solutions provides visibility, detection, and prevention of threats like IcedID at every stage of the attack.


Malicious documents SHA256