During our threat intelligence source monitoring operations, we spotted a new sophisticated malware implant, dubbed JsOutProx, that seems to be unrelated to mainstream cyber weapons. In fact, the recovered sample raised many interrogatives into the malware research community due to the extensive usage of obfuscation anti-reverse techniques, hardening the investigative efforts.
For this reason, we decided to dig into this piece of malware and figure out its inner secrets, uncovering a modular architecture with advanced offensive capabilities, such as the presence of functionalities able to deal with multi-factor authentication (MFA).
|Brief Description||Malicious JS file|
Table 1. JsOutProx Sample information
The first line of this file embeds a huge array of Base64 encoded elements, but its raw decoding led only to other incomprehensible data, evidencing the presence of a more complex layer of protection.
Navigating the code, we identified a series of instructions resembling a sort of initialization that grabbed our attention. The function “t_ey” is used as
This initial code cleanup revealed interesting information such as some of the static configuration initialized during the initial malware execution stages. Among this info, we recovered also its remote C2 address 184.108.40.206, operated by “ServeTheWorld”, a Norwegian provider renting his servers in Oslo, and a particular tag reporting the name “JsOutProx”. Extremely characteristic.
Continuing to analyze the code, we reconstructed the approach used by the attacker to obfuscate the payload: all the necessary information has been encrypted, splitted, and then encoded in Base64 chunks stored into different structures named as “ta”, “t_ep”, “t_eq”.
As anticipated before, thanks to the decoding routine “t_ey” it is possible to retrieve at runtime the cleartext code to reify the structure of the malware, stored in the “t_fT” object visible in the next figure.
The structure contains objects and functions used by the malware to pursue its malicious actions. In many cases we noticed a naming correspondence between couples of objects, for example between “Outlook” and “OutlookPlugin”, or “Proxy” and “ProxyPlugin” objects. This indicates the malware has a modular structure containing specific plugins able to perform a wide range of actions, such as exfiltrate data by populating the associated object. For example, the “OutlookPlugin” is able to steal information about emails and contacts and it does that by filling the Outlook object shown in the previous figure.
Each plugin embeds an obfuscated function named “receive”, which has the purpose to perform the specific action. This function name is constant and represent a sort of common interface between malware modules.
Once created the main structure, the first function ran by the malware is “init”. It is designed to create an identification string for the victim machine, gathering from system information and storing them into the “t_fT[“ID”]” variable.
As visible the next picture, the identification string is composed using the computer name, username, OS version. Then, an additional suffix will be appended to it containing the current action the malware is performing. In this specific case it is “ping”, a sort of heartbeat to make sure the command and control services is up and running.
Then, the malware moves in an endless loop in which it invokes the “receive” function every 5 seconds. Substantially, this function is an interface that allows the attacker to interact with the implant.
Summing up, the analyzed version of the implant supports the following actions:
|upd||Update the implant|
|rst||Restart the implant|
|l32||Start another process with the same script|
|dcn||Kill the implant|
|rbt||Reboot the machine|
|shd||Shutdown the machine|
|evb||Execute VisualBasic code|
|sdn||Load a .NET dll|
|uis||Uninstall the implant|
|ins||Install the implant|
|int.g||Send the sleep time to C2|
|int.s||Update the sleep time|
Table 2. Supported commands.
Additionally, the default case of the main loop structure includes a series of IF-Else evaluating the received string in order to check for specific prefixes. Each prefix is associated to a plugin module.
Analyzing this selection chain was possible to reconstruct the full list of supported plugin for this version of the implant.
Table 3. Malware’s plugins.
At this point, all the attention moves on to understand what the functions and the plugins do and how they work. So, we provide, in the next paragraphs, a brief view of the most interesting functions and plugins.
By the way, to ensure its survival after reboot, the malware writes a copy of the initial JS script into the “%appdata%” and “%temp%” folder, setting the persistence on the system through the popular “HKCU\Software\Microsoft\Windows\CurrentVersion\Run” registry key.
The number of supported plugin indicates the malware is pretty mature and extremely dangerous. Some of these supported plugins encode standard functionalities of many RAT and recon malware, but some of them hides interesting and even uncommon features.
The ProcessPlugin is able to manipulate other processes running in the system. It can kill them by PID and by name, create new processes through WSH or WMI and also collect a memory dump of a specific process. This is an uncommon feature for a common malware, it might indicate operators behind its command and controls could have analytical skills in order to investigate the surrounding environment, or also exfiltrate data through memory scraping.
The DnsPlugin handles the machine’s DNS configuration. It can send to the C2 the current configuration and also set a new one, just like visible in the following screen.
This plugin is particularly interesting. The name refers to the exfiltration of some type of token. However, the plugin is specifically designed for the theft of SymantecVIP One Time Password, the multifactor authentication technology used in enterprise grade Single-Sign-On services adopted in many corporate environments.
Instead, the OutlookPlugin weaponize the implant with common information stealing capabilities enabling the attackers to gather account information and contact list.
The interesting part of this plugin is that it is the only one info-stealing plugin embedded in the malware.
Commodity malware typically include support to multiple email clients, web browsers and ftp clients, but this one just handle Outlook: the de facto standard email client installed in the majority of enterprise environment.
Another interesting functionality is provided by the PromptPlugin. It empowers the attacker to present his victim a custom message prompt provided by the command and control server.
Also, digging the JS code, a method called “hasDotnet” made us wondering about the existence of additional .NET artifacts which may hide the implementation of some of the above-mentioned functionalities. Moreover, this method contains references to the download and the memory loading of an instance of a “DotUtil” module.
Dynamically analyzing the malware, we were able to intercept the download of a .NET DLL called “libDotJs.dll”. Analyzing it, we discover a complete mapping between the functions implemented in it and the ones declared into “DotUtil” component.
The complexity and the engineering grade of this threat, dubbed JsOutProx, are an element of uniqueness in the current panorama. It is a toolkit with peculiar remote access capabilities. From an architectural point of view, JsOutProx contains all the function prototypes inside the core engine and could be remotely extended at run time.
Another relevant aspect of JsOutProxy is the capability to deal with SymantecVIP technology. This led us to think that, this implant, has been designed to hit High-Value targets. Also, this new threat appears emerging during these days, it has never been publicly seen before this December and it is probably still under development.
Additional details, including Indicator of Compromise (IoC) and Yara Rules, are reported in the analysis published by Yoroi-Cybaze.
|[adrotate banner=”9″]||[adrotate banner=”12″]|