ESET takes part in global operation to disrupt Zloader botnets | WeLiveSecurity

ESET has collaborated with partners Microsoft’s Digital Crimes Unit, Lumen’s Black Lotus Labs, Palo Alto Networks Unit 42, and others in an attempt to disrupt known Zloader botnets. ESET contributed to the project by providing technical analysis, statistical information, and known command and control server domain names and IP addresses.

Zloader started life as a banking trojan, but lately evolved to become a distributor of several malware families, including various ransomware families.

The coordinated disruption operation targeted three specific botnets, each one using a different version of the Zloader malware. ESET researchers helped with identification of 65 domains that had been used by these botnet operators recently and that had been taken over for this disruption operation to be effective. On top of that, Zloader bots rely on a backup communication channel that automatically generates unique domain names that can be used to receive commands from their botmasters. This technique, known as a domain generation algorithm (DGA), is used to generate 32 different domains per day, per botnet. To make sure that the botnet operators cannot use this side channel to regain control of their botnets, additional 319 already registered domains generated by this algorithm were taken over and the working group is also taking measures to block registration of DGA domains possibly generated in the future. Microsoft’s investigation also identified Denis Malikov as a co-author of a malicious component used by the operators of one of the botnets.

Background

Zloader is one of the many banking trojan malware families heavily inspired by the famous Zeus banking trojan, whose source code was leaked in 2011. Many research papers have been published about this malware already, with the latest one from Malwarebytes and HYAS being the most detailed from the technical point of view.

This blogpost won’t focus on deep technical aspects of the trojan, but rather will cover the details of its operation and infrastructure.

The first version (1.0.0.0) of Zloader that we were able to find was compiled on November 9^th 2019, the same day it was announced and advertised in underground forums under the name “Silent Night”. ESET researchers have been closely monitoring its activity and evolution ever since then, giving us great insight into Zloader’s mode of operation and its infrastructure.

Throughout Zloader’s existence, we have analyzed about 14,000 unique samples via our automatic tracking system, which helped us to discover more than 1,300 unique C&C servers. In March 2020, Zloader implemented a domain generation algorithm (DGA) that allowed us to discover about 300 additional active domains registered by Zloader operators and used as C&C servers.

We have seen a couple of peaks in Zloader’s popularity among threat actors, mainly during its first year of existence, but its use began declining during 2021 with only a couple of actors left using it for their malicious intents. This may, however, change in the future as we have already seen version 2.0 samples in the wild (compiled in July 2021). Our findings show that these were just test builds, but we will be closely monitoring this new activity and its evolution. Due to low prevalence and the nature of this new version, all the following information applies to Zloader version 1.x.

As already mentioned, Zloader, similar to other commodity malware, is being advertised and sold on underground forums. When purchased, affiliates are given all they need to set up their own servers with administration panels and to start building their bots. Affiliates are then responsible for bot distribution and maintaining their botnets.

As you can see in Figure 1, we have observed Zloader infestations and campaigns in many countries with North America being the most targeted.

Zloader has been used by various affiliate groups and each of them has used a different approach for the malware’s distribution, including:

• RIG exploit kit
• COVID-19-themed spam emails with malicious Microsoft Word documents attached
• Variants of a fake invoice spam emails with malicious XLS macros
• Misuse of Google Ads

The development of the latest distribution methods will be covered in the next sections.

Zloader internals

Zloader has a modular architecture, downloading and utilizing its modules as needed. Supported Zloader modules are displayed in Table 1 and Table 2.

Table 1. Overview of malicious modules used by Zloader

Table 2. Legitimate tools abused by Zloader to support its malicious tasks

Zloader’s first component is a loader that is used to download or load (if already downloaded) the core module. This core module is then responsible for downloading and loading additional modules and performing its own malicious tasks.

Zloader’s notable features are:

• Ability to steal various data from browsers and Microsoft Outlook, steal cryptocurrency wallets
• Keystroke logging
• HiddenVNC support to allow the operator to remotely control compromised systems
• Support for Zeus-like webinjects, form grabbing and form screenshotting
• Arbitrary command execution (e.g., download and execute other malware)

All communication between bots and their C&C servers is performed over HTTP/HTTPS, and regardless of which is used the data is encrypted using RC4. Some of the data is additionally encrypted using an XOR-based algorithm known as “Visual Encrypt”. The RC4 key is unique for each affiliate as described in the next section. Figure 2 shows a bot’s static configuration. It contains a list of up to ten hardcoded C&C URLs along with other important data for communication – such as the botnetID to help the operator easily filter data from different campaigns, the signature for communications verification, etc. A bot’s C&C list can be easily updated by issuing a command from the operator’s administration panel if needed.

Figure 2. Zloader’s static configuration

If none of the hardcoded servers responds, a Zloader bot can use its DGA as a fallback mechanism. Every day, a list of 32 new domains unique for every affiliate is generated based on the current day retrieved by GetLocalTime function. Generated URLs have the format https://<20_random_lowercase_ASCII_letters>.com/post.php

Botnet infrastructure and affiliates

The RC4 encryption key used in botnet communication is unique for every affiliate and tied to the affiliate’s administration panel installation. This uniqueness gives us the opportunity to cluster Zloader samples and track affiliates’ distribution methods and the evolution of their campaigns.

Since the beginning of our tracking, we have observed more than 25 different RC4 keys. It is worth noting that some of these affiliates were active for a very short period — some of them were probably just testing Zloader’s features. It is also possible that some operators just redeployed their administration panel installation at some point and continued their operation with a new RC4 key. A timeline of notable affiliate activity, as well as various Zloader version release dates, can be seen in Figure 3.

Figure 3. Activity of some of the notable affiliates

As can be seen in Figure 5, from October 2020, most Zloader activity was due to only two affiliates. We can distinguish them by their RC4 keys – 03d5ae30a0bd934a23b6a7f0756aa504 and dh8f3@3hdf#hsf23

We cover these two affiliates’ activities in the next two sections.

dh8f3@3hdf#hsf23

This affiliate was active under this particular RC4 key starting in June 2020. The first Zloader version it used was 1.3.27.0 and then closely followed the newest version available up until the latest available Zloader version to this date – 1.8.30.0. However, its activity started to decline in the second half of 2021 and we haven’t seen any new activity of this botnet since late November 2021.

One of the most interesting activities of this affiliate is that it used Zloader’s ability to deploy arbitrary payloads to distribute malicious payloads to its bots. Most notably, it spread various ransomware families such as DarkSide, as highlighted by this research from Guidepoint Security. However, the botmaster did not deploy ransomware to all of their bots; they deployed this type of malware mostly on systems belonging to corporate networks. When installed on a system, Zloader gathers various information about the network its compromised host belongs to. This allows botnet operators to pick specific payloads depending on the victim’s network.

This affiliate was spreading their malicious Zloader samples mostly through spam emails with malicious documents attached to them. The Zloader static configuration contains a botnetID, allowing the botmaster to cluster different bots in different sub-botnets. The most prevalent botnetIDs for this affiliate in the last year of its operation were nut and kev.

This operator was also a bit more security aware compared to other Zloader customers and used a tiered architecture for their C&C servers. Typically, a simple proxy script was planted on an often legitimate but compromised website and it was used for tier1 C&C URLs in their bots. This script simply forwards all HTTP/HTTPS traffic from the bot onto the tier2 server, keeping the location of the real administration panel installation secret.

Besides using Zloader as an entry point for ransomware attacks, this affiliate also used Zloader’s adversary-in-the-browser (AitB) capabilities to steal victim information and alter the content of various financial institutions and e-commerce websites based in the USA and Canada.

03d5ae30a0bd934a23b6a7f0756aa504

This affiliate has been using Zloader since its early versions and is still active as of today. Despite the latest available version of Zloader being 1.8.30.0, this affiliate has stuck with version 1.6.28.0 since its release in October 2020. We can only speculate as to the reasons behind this. One hypothesis is that this affiliate did not pay to extend their support coverage for Zloader and thus does not have access to later versions.

The operator of this botnet used to depend solely on C&C domains generated by Zloader’s DGA and did not update their bots with a new C&C list for more than a year, meaning that all hardcoded C&C servers in their bots were inactive for a long time. This changed in November 2021 when this affiliate updated their bots with a list of new C&C servers and also updated the static configuration of newly distributed binaries to reflect this change. This effort was probably motivated by the fear of losing access to their botnet should anyone register and sinkhole all future DGA-generated domains for this actor.

Figure 4 shows the administration panel login page which was installed directly on the C&C server hardcoded in the bot’s static configuration.

Figure 4. Administration panel login page

Some notable botnetIDs used by this operator were: personal, googleaktualizacija and more recently return, 909222, 9092ti and 9092us.

Through analysis of the webinjects downloaded by the bots in this affiliate botnet, the operator’s interests are very broad. They are apparently interested in gathering victim’s login credentials and other personal data from various financial institution websites (banks, stock trading platforms, etc.), e-commerce sites (such as Amazon, Best Buy, Walmart), cryptocurrency exchanges and even various online platforms such as Google and Microsoft. Particular focus was put on customers of financial institutions from the USA, Canada, Japan, Australia and Germany.

Additional to the login credential harvesting, this affiliate also used Zloader to distribute various malware families such as the infostealer Raccoon.

Distribution

This threat actor uses various means to spread Zloader with misusing Google Ads and bogus adult sites being their latest distribution methods of choice.

Starting in October 2020, fake adult sites started to push to their visitors malicious payloads posing as a Java update in an MSI package (with filename JavaPlug-in.msi), supposedly required to watch the requested video. This fake Java update package typically contained a downloader that downloaded Zloader itself as the final payload. Since April 2021, this scheme has been enhanced by adding a script to disable Microsoft Defender to further increase the chances of successfully compromising the system.

In June 2021, this affiliate also started to promote packages typically used in corporate environments. When internet users searched for a popular application to download, such as Zoom or TeamViewer, they might have been presented with a fake download site promoted via a Google Ad that tried to trick them into downloading a malicious package posing as the app they were searching for. This distribution method not only installed Zloader but could also install other potentially malicious tools, notably if the compromised system was part of an Active Directory domain. Notorious Cobalt Strike Beacon and Atera Agent were seen to be installed in such cases. These tools could grant the attacker complete control of the compromised system and may result in stealing of sensitive company data, installation of other malware such as ransomware and other malicious activity incurring significant losses for the company.

Figure 5 shows the logic to check if a system belongs to a domain. As seen below, Cobalt Strike Beacon is installed if the list of the system’s trusted domains is non-empty.

Figure 5. PowerShell script responsible for Cobalt Strike Beacon installation

The latest iteration of this distribution method relied heavily on the aforementioned Atera Agent, which was usually downloaded from bogus adult sites. An example of what a visitor would see is shown in Figure 6.

Figure 6. Fake adult site luring users into downloading Atera remote management tool

Atera Agent is a legitimate “remote monitoring and management” solution used by IT companies to administer their customers’ systems. One of its features – remote script execution – was used in this campaign to deliver Zloader payloads and other malicious helper files. The purpose of these helper files was to support the installation process by executing specific tasks such as privilege escalation, execution of further sample, disabling of Windows Defender, etc.

These tasks were usually achieved via simple BAT files, but it is worth mentioning that attackers also exploited a known digital signature verification vulnerability to use legitimate, signed Windows executable files with malicious VBScripts appended to the end of the file, where the signature section is located (see Figure 7). For the PE file to remain valid, attackers also need to alter the PE header to alter the signature section length and checksum. This alteration of the file’s content does not revoke the validity of its digital signature during the verification process because the modified content is exempted from the verification process. Thus, the file’s new malicious content may therefore stay off the radar. This vulnerability is described, for example, in CVE-2012-0151 or CVE-2013-3900, and also in this blogpost by Check Point Research. Its fix is unfortunately disabled by default in Windows, and therefore, it still can be misused by attackers in a large number of systems.

Figure 7. Example of a script appended to the PE file signature section

In the recent campaign, a Ursnif trojan was sometimes installed instead of Zloader, showing that this affiliate group does not rely on a single malware family but has more tricks up its sleeve. A typical scenario of this distribution method is displayed in Figure 8.

Figure 8. Typical distribution method using Atera Agent

Closing remarks

We  relentlessly continue to track threats that are used to spread ransomware, which is an ongoing threat to internet security. As Zloader is available in underground forums, ESET Researchers will monitor any new activity tied to this malware family, following this disruption operation against its existing botnets.

IoCs

Samples

Network

Domains and URLs used in distribution

• https://endoftheendi[.]com
• https://sofftsportal[.]su
• https://pornokeyxxx[.]pw
• https://porno3xgirls[.]website
• https://porno3xgirls[.]space
• https://porno3xgirls[.]fun
• https://porxnoxxx[.]site
• https://porxnoxxx[.]pw
• https://pornoxxxguru[.]space
• https://helpdesksupport072089339.servicedesk.atera[.]com/GetAgent/Msi/?customerId=1&[email protected]
• https://helpdesksupport350061558.servicedesk.atera[.]com/GetAgent/Msi/?customerId=1&[email protected]
• https://clouds222[.]com
• https://teamworks455[.]com
• https://commandaadmin[.]com
• https://cmdadminu[.]com
• https://checksoftupdate[.]com
• https://datalystoy[.]com
• https://updatemsicheck[.]com

Latest Zloader C&C servers

• https://asdfghdsajkl[.]com/gate.php
• https://lkjhgfgsdshja[.]com/gate.php
• https://kjdhsasghjds[.]com/gate.php
• https://kdjwhqejqwij[.]com/gate.php
• https://iasudjghnasd[.]com/gate.php
• https://daksjuggdhwa[.]com/gate.php
• https://dkisuaggdjhna[.]com/gate.php
• https://eiqwuggejqw[.]com/gate.php
• https://dquggwjhdmq[.]com/gate.php
• https://djshggadasj[.]com/gate.php

URLs used to download arbitrary malware

• https://braves[.]fun/racoon.exe
• https://endoftheendi[.]com/12.exe

Domains used in recent Zloader’s Webinjects attacks

• https://dotxvcnjlvdajkwerwoh[.]com
• https://aerulonoured[.]su
• https://rec.kindplanet[.]us

MITRE ATT&CK techniques

This table was built using version 10 of the MITRE ATT&CK framework.