Jan 3, 2022
0 0

Stantinko botnet adds cryptomining to its pool of criminal activities

Written by

ESET researchers have discovered that the criminals behind the Stantinko botnet are distributing a cryptomining module to the computers they control
The operators of the Stantinko botnet have expanded their toolset with a new means of profiting from the computers under their control. The roughly half-million-strong botnet – known to have been active since at least 2012 and mainly targeting users in Russia, Ukraine, Belarus and Kazakhstan – now distributes a cryptomining module. Mining Monero, a cryptocurrency whose exchange rate has oscillated in 2019 between US$50 and US$110, has been the botnet’s monetizing functionality since at least August 2018. Before that, the botnet performed click fraud, ad injection, social network fraud and password stealing attacks.
In this article, we describe Stantinko’s cryptomining module and provide an analysis of its functionality.
This module’s most notable feature is the way it is obfuscated to thwart analysis and avoid detection. Due to the use of source level obfuscations with a grain of randomness and the fact that Stantinko’s operators compile this module for each new victim, each sample of the module is unique.
We will describe the module’s obfuscation techniques and offer, in a separate article for fellow malware analysts, a possible approach to deal with some of them.
Since Stantinko is constantly developing new and improving its existing custom obfuscators and modules, which are heavily obfuscated, it would be backbreaking to track each minor improvement and change that it introduces. Therefore, we decided to mention and describe only what we believe are significant adjustments in comparison with earlier samples relative to the state in which the module is to be described. After all, we intend just to describe the module as it currently is in this article.
Stantinko’s cryptomining module, which exhausts most of the resources of the compromised machine by mining a cryptocurrency, is a highly modified version of the xmr-stak open-source cryptominer. All unnecessary strings and even whole functionalities were removed in attempts to evade detection. The remaining strings and functions are heavily obfuscated. ESET security products detect this malware as Win{32,64}/CoinMiner.Stantinko.
CoinMiner.Stantinko doesn’t communicate with its mining pool directly, but via proxies whose IP addresses are acquired from the description text of YouTube videos. A similar technique to hide data in descriptions of YouTube videos is used by the banking malware Casbaneiro. Casbaneiro uses much more legitimate-looking channels and descriptions, but for much the same purpose: storing encrypted C&Cs.
The description of such a video consists of a string composed of mining proxy IP addresses in hexadecimal format. For example, the YouTube video seen in Figure 1 has the description “03101f1712dec626“, which corresponds to two IP addresses in hexadecimal format – 03101f17 corresponds to 3.16.31[.]23 in decimal dotted-quad format, and 12dec626 is 18.222.198[.]38. As of the time of writing, the format has been slightly adjusted. The IP addresses are currently enclosed in “!!!!”, which simplifies the very process of parsing and prevents possible changes of the YouTube video HTML structure turning the parser dysfunctional.
Figure 1. Example YouTube video whose description provides an IP address for the module’s communication with the mining pool
In earlier versions, the YouTube URL was hardcoded in the CoinMiner.Stantinko binary. Currently the module receives a video identifier as a command line parameter instead. This parameter is then used to construct the YouTube URL, in the form https://www.youtube.com/watch?v=%PARAM%. The cryptomining module is executed either by Stantinko’s BEDS component, or by rundll32.exe via a batch file that we have not captured, with the module loaded from a local file system location of the form %TEMP%%RANDOM%%RANDOM_GUID%.dll.
We informed YouTube of this abuse; all the channels containing these videos were taken down.
We have divided the cryptomining module into four logical parts, which represent distinct sets of capabilities. The main part performs the actual cryptomining; the other parts of the module are responsible for additional functions:
At the very core of the cryptomining function lies the process of hashing, and communication with the proxy. The method of obtaining the list of mining proxies is described above; CoinMiner.Stantinko sets the communication with the first mining proxy it finds alive.
Its communication takes place over TCP and is encrypted by RC4 with a key consisting of the first 26 characters of the number pi (including the decimal separator, hardcoded in the string “3,141592653589793238462643“) and then base64 encoded; the same key is used in all samples we have seen.
The code of the hashing algorithm is downloaded from the mining proxy at the beginning of the communication and loaded into memory – either directly or, in earlier versions, from the library libcr64.dll that is first dropped onto the disk.
Downloading the hashing code with each execution enables the Stantinko group to change this code on the fly. This change makes it possible, for example, to adapt to adjustments of algorithms in existing currencies and to switch to mining other cryptocurrencies in order, perhaps, to mine the most profitable cryptocurrency at the moment of execution. The main benefit of downloading the core part of the module from a remote server and loading it directly into memory is that this part of the code is never stored on disk. This additional adjustment, which is not present in earlier version, is aimed at complicating detection because patterns in these algorithms are trivial for security products to detect.
All instances of Stantinko’s cryptomining module we’ve analyzed mine Monero. We deduced this from the jobs provided by the mining proxy and the hashing algorithm. For example, Figure 2 is a job sent by one of the proxies.


Figure 2. Example mining job received from a mining pool proxy
We analyzed the hashing algorithm used and found that it was CryptoNight R. Since there are multiple cryptocurrencies that use this algorithm, its recognition alone isn’t sufficient; it just shortens the list. One can see in the provided job that the height of the blockchain was 1815711 at the time, so we had to find currencies using CryptoNight R with this height on dedicated block explorers which lead us to Monero. Dissecting the string 0b0bbfdee1e50567042dcfdfe96018227f25672544521f8ee2564cf8b4c3139a6a88c5f0b32664000000a1c8ee5c185ed2661daab9d0c454fd40e9f53f0267fe391bdb4eb4690395deb36018 reveals that the hash of the previous block (67042dcfdfe96018227f25672544521f8ee2564cf8b4c3139a6a88c5f0b32664) and timestamp (1555590859) indeed fits into Monero’s blockchain at the height of 1815711. One can find the structure of the blob by examining its generator function in the source code of Monero . The generator function exposes another structure called a block header which contains both the hash of the previous block and timestamp.
Unlike the rest of CoinMiner.Stantinko, the hashing algorithm isn’t obfuscated, since obfuscation would significantly impair the speed of hash calculation and hence overall performance and profitability. However, the authors still made sure not to leave any meaningful strings or artifacts behind.
The malware enumerates running processes searching for other cryptominers. If any competitors are found, Stantinko suspends all their threads.
CoinMiner.Stantinko considers a process to be a cryptominer if its command line contains a particular string, or a combination, which vary from sample to sample; for example:
These strings refer to the following legitimate cryptominers: https://github.com/pooler/cpuminer, https://minergate.com/, https://github.com/xmrig, and even https://github.com/fireice-uk/xmr-stak – which, interestingly, is the very miner this Stantinko module is based on. The strings also lead to various uninteresting malware samples containing cryptomining functionality.
Of interest is that the Stantinko operators are known to have tried to get rid of competing code in the past. However, they relied on the legitimate AVZ Antiviral Toolkit fed with a script written in its built-in scripting language for this task.
CoinMiner.Stantinko temporarily suspends mining if it detects there’s no power supply connected to the machine. This measure, evidently aimed at portable computers, prevents fast battery draining … which might raise the user’s suspicion.
Also, it temporarily suspends mining if a task manager application (a process named procexp64.exe, procexp.exe or taskmgr.exe) is detected running.
The malware also scans running processes to find security software and again task managers. It calculates the CRC-32 of the process’s name and then checks it against a hardcoded list of CRC-32 checksums, which is included in the Appendix. In general this technique can help evade detection, since the process names of those security products are not included in the binary – adding a bit more stealth by not containing the process names directly. It also makes it harder for analysts to find out what the malware authors are after because one has to crack these hashes, which is technically the same problem as password cracking. However, using a list of known process names is usually sufficient to determine the exact names.
Should a CRC-32 match be found, the CRC is written to a log file (api-ms-win-crt-io-l1-1-0.dll). The log file is presumably exfiltrated later by some Stantinko component that we have not seen, since there’s no other functionality related to it in this module.
Besides its cryptomining features, CoinMiner.Stantinko is notable also for its obfuscation techniques aimed at avoiding detection and thwarting analysis. Some of those techniques are unique and we will describe them in detail in a follow-up article.
Our discovery shows that the criminals behind Stantinko continue to expand the ways they leverage the botnet they control. Their previous innovations were distributed dictionary-based attacks on Joomla and WordPress web sites aimed at harvesting server credentials, probably with the goal of selling them to other criminals.
This remotely configured cryptomining module, distributed since at least August of 2018 and still active at the time of writing, shows this group continues to innovate and extend its money-making capabilities. Besides its standard cryptomining functionality, the module employs some interesting obfuscation techniques that we will disclose, along with some possible countermeasures, in an upcoming article.
A full list of more than 1,000 hashes is available from our GitHub repository.
• 3.16.150[.]123
• 3.16.152[.]201
• 3.16.152[.]64
• 3.16.167[.]92
• 3.16.30[.]155
• 3.16.31[.]23
• 3.17.167[.]43
• 3.17.23[.]144
• 3.17.25[.]11
• 3.17.59[.]6
• 3.17.61[.]161
• 3.18.108[.]152
• 3.18.223[.]195
• 13.58.182[.]92
• 13.58.22[.]81
• 13.58.77[.]225
• 13.59.31[.]61
• 18.188.122[.]218
• 18.188.126[.]190
• 18.188.249[.]210
• 18.188.47[.]132
• 18.188.93[.]252
• 18.191.104[.]117
• 18.191.173[.]48
• 18.191.216[.]242
• 18.191.230[.]253
• 18.191.241[.]159
• 18.191.47[.]76
• 18.216.127[.]143
• 18.216.37[.]78
• 18.216.55[.]205
• 18.216.71[.]102
• 18.217.146[.]44
• 18.217.177[.]214
• 18.218.20[.]166
• 18.220.29[.]72
• 18.221.25[.]98
• 18.221.46[.]136
• 18.222.10[.]104
• 18.222.187[.]174
• 18.222.198[.]38
• 18.222.213[.]203
• 18.222.253[.]209
• 18.222.56[.]98
• 18.223.111[.]224
• 18.223.112[.]155
• 18.223.131[.]52
• 18.223.136[.]87
• 18.225.31[.]210
• 18.225.32[.]44
• 18.225.7[.]128
• 18.225.8[.]249
• 52.14.103[.]72
• 52.14.221[.]47
• 52.15.184[.]25
• 52.15.222[.]174
CRC-32 checksums checked by CoinMiner.Stantinko and the filenames they equate to are listed below.


Article Categories:

Comments are closed.