- Why Cleafy
- Solution
- Intelligence
- Resources
- Company
- Get in touch
Get in touch
At the end of May 2024, a new Android RAT appeared in Cleafy’s telemetries.
Due to the lack of information and the absence of a proper nomenclature for this malware family, we decided to dub it BingoMod to track it inside our Threat Intelligence taxonomy. This nomenclature is based on the malware's core component, known at an early stage as “ChrUpdate” but later renamed “BingoMod”.
BingoMod belongs to the modern RAT generation of mobile malware, as its remote access capabilities allow Threat Actors (TAs) to conduct Account Takeover (ATO) directly from the infected device, thus exploiting the On Device Fraud (ODF) technique. This consolidation of this technique has already been seen recently by other banking trojans, such as Medusa, Copybara, and Teabot.
These techniques have several advantages: they require less skilled developers, expand the malware's target base to any bank, and bypass various behavioural detection countermeasures put in place by multiple banks and financial services. However, this advantage does not come for free, one of the drawbacks of this technique relies on a live operator that is required to insert and authorise a money transfer, which implicitly means lowering its scale factor.
BingoMod is similar to the Brata's operation model in using device wiping after a successful fraudulent transfer. This self-destruction mechanism is designed to eradicate any trace of BingoMod's activity on the infected device, effectively hindering forensic analysis and making it more challenging for researchers to identify and attribute incidents. This tactic is relatively rare in the Android landscape, suggesting that the developers of BingoMod could be aware of Brata's methods and have incorporated them into their methodology.
Moreover, it's also worth mentioning that this sample is in its early stage of development, and it's still hard to predict which direction will be taken. However, the developers’ commitment to attempting obfuscation techniques underlines their intention to pursue a more opportunistic approach than a tailored one already seen in malware like SharkBot or Gustuff.
The following table represents a summary of the TTP behind BingoMod campaigns:
All detected samples provided in the Appendix are disguised as legitimate security tools to protect the device.
As previously mentioned, the malicious app is distributed via smishing and often masquerades as a legitimate antivirus application. After installation on the victim's device, BingoMod prompts the user to activate Accessibility Services, disguising the request as necessary for the app to function correctly. If the user grants the requested permissions, the APK begins to unpack itself, executing its malicious payload. Once the operation is completed, the apps still lock out the user from the main screen to collect device information and set up the C2 communication channel.
After activation, BingoMod's background functions act, aiming to provide sensitive data to the actors behind the malware. In detail, two features typical of banking Trojans are used:
As mentioned earlier, BingoMod's main objective is to initiate money transfers directly from compromised devices (ODF). Therefore, the malware implements several remote control functionalities. To do this, BingoMod establishes a socket-based connection with the command and control infrastructure (C2) to receive commands that TAs want to perform on the compromised device.
The malware provides around 40 remote control functions, among the most relevant are indeed related to the real-time screen control that is implemented in the following way:
For this purpose, BingoMod uses two separate communication channels: a socket-based channel for command transmission (in the case of VNC start/stop) and an HTTP-based channel for image transfer.
To better illustrate this process, we created a simulated C2 infrastructure. This setup includes a socket-based C2 server for remote control and an HTTP-based "VNC" server to capture and display real-time screenshots the infected device sends.
On the malware side, the VNC routine abuses Android's Media Projection API to obtain real-time screen content. Once received, this is transformed into a suitable format and transmitted via HTTP to the TAs' infrastructure. An exciting feature of the routine is leveraging Accessibility Services to impersonate the user and enable the screen-casting request, exposed by the Media Projection API.
Once the VNC-like routine is activated, TAs can interact with the device using dedicated commands. These include, for example, opening a specific application (<LAUNCH>), moving to a particular area on the screen (<MOVEAT>), clicking a particular area (<CLICKAT>) or writing in a particular text box (<SETTEXT>).
In addition to real-time screen control, the malware shows phishing capabilities through Overlay Attacks and fake notifications. Unusually, overlay attacks are not triggered when specific target apps are opened but are initiated directly by the malware operator. Still, in the context of phishing, TAs can also send SMS messages from the compromised device; this functionality can be used to spread the malware further.
Finally, the malware implements some security measures to improve its resilience: it hinders editing system settings, especially one regarding the malware itself, and can block the activity of specified applications set by the actors via a dedicated command. If this is not enough, BingoMod can also uninstall arbitrary applications. For instance, this feature can be used to prevent security apps from detecting the presence of the malware itself.
As mentioned earlier, BingoMod's most notable security measure is its ability to wipe the device remotely with a dedicated command. This feature can be implemented by BingoMod when it is a device administrator and is typically executed after a successful fraud. However, this functionality is limited to the device's external storage only, so we speculate that the complete wipe is performed by TAs directly from the device's system settings, leveraging BingoMod's remote access capabilities.
The entire BingoMod command set is provided in the appendix.
It’s worth considering that from the first sample until now, it has been possible to observe that developers are in an “experimental” phase, mainly on app obfuscation and packing process that aims to reduce its detection against AV solutions instead of equipping their code with advanced capabilities. This change is immediately observable using detection engines from VirusTotal, showing that early campaigns were easily detected, whereas the recent one dropped their detection rate.
As mentioned, functions and classes stayed mostly the same over time. However, the obfuscation employed lowers the overall detection rate. The figure below gives a code example that refers to the hiddenVNC procedure, where the malware is sending a screenshot to the C2, simulating a live stream.
In the upper part of this image, we can see an early version of the sample. Aside from some variable renaming, the overall code is understandable. However, the same function appears to be heavily obfuscated using code-flattening and string obfuscation techniques. After tweaking the code to resolve the switches, the overall structures are the same, strengthening the hypothesis that TAs rely on obfuscation over malware complexity.
Moreover, it’s worth comparing different code versions to analyse their changes when discussing malware evolution. As shown in Figure 11, there were no significant changes in the overall structure and functionality. However, an asynchronous callback mechanism has been introduced in the PingUtil class to send "alive" signals to the C2 server, giving information about the bot's status.
Another evidence confirming its developmental stage is the amount of log information left within the code. It is worth mentioning that although some comments have been deleted in future releases, some still refer to older package names. For instance, in release 1.4.3b, references to package names from the first version (1.0) are still visible.
Another interesting element that emerged during the BingoMod investigation is related to target devices, which include three languages: English, Romanian, and Italian, here disguised as com.
However, it is possible to observe that the Romanian language is mixed with English words. For the Italian language, it has also been observed with a few typos. Those strings and the general code could be another sign that those versions are still in their “debug” phase.
Details retrieved during the investigation led us to several speculations about the developers. For instance, some comments are in Romanian, but these comments have been removed in more recent versions. This suggests that the TAs have evolved, possibly incorporating developers from different countries, or they are trying to reduce indicators that can lead further LEA investigation to a specific region.
Moreover, a recent campaign involving version 1.4.3b was uploaded to VirusTotal from an IP address geolocating to the Romanian region. While this behaviour is plausible, it strengthens suspicions about active campaigns and areas involved in their attack.
BingoMod shows relatively straightforward functionalities commonly found in most contemporary RAT, such as HiddenVNC for remote control and SMS suppression to intercept and manipulate communication and logging user interactions to steal sensitive data. The emphasis on obfuscation and unpacking techniques suggests that the developers may lack the sophistication or experience of more advanced malware authors. This is evident in their reliance on basic obfuscation methods and the minimalistic approach to developing the malware’s functionalities. It is plausible that the developers aimed to quickly produce a functional piece of malware, prioritising speed over complexity.
One notable aspect of this malware is its device-wiping capability, triggered after a fraudulent transaction. This behaviour is reminiscent of the Brata malware, which also employed device-wiping to cover its tracks and hinder forensic analysis. However, the simplicity and rudimentary nature of the code suggests that this feature is more of an easy exit strategy rather than an indication of any direct lineage or connection to Brata. This action further supports the theory that the developers opted for straightforward and efficient methods to achieve their goals without delving into more complex development practices. Compared to more sophisticated threats such as Sharkbot, which incorporates an Automatic Transfer System (ATS) to automate fraudulent transactions, this malware needs to improve in terms of advanced capabilities.
The lack of automated components implies that scaling this approach would be challenging, as it requires direct operator interaction to perform Account Takeover (ATO) activities. This direct interaction often involves manipulating the victim or their device in real time, limiting the scale of potential attacks and exposing the operators to a higher risk of detection. The reliance on manual intervention underscores the limited technical capabilities behind this malware and indicates that it operates within a more traditional fraud paradigm. Interestingly, the current trend among mobile banking trojans, exemplified by threats like Copybara and Medusa, focuses on On-Device Fraud (ODF) through ATO. This shift highlights a broader trend within the malware landscape where the emphasis is placed on exploiting the device to conduct fraudulent activities rather than developing highly automated systems. The analysed malware aligns with this trend, focusing on direct interaction with the victim's device to carry out its malicious objectives. This approach, while less sophisticated, still poses significant risks to end-users and financial institutions due to the potential for substantial economic loss and the disruption of personal data security.
