From Caribbean shores to your devices: analyzing Cuba ransomware

Introduction

Knowledge is our best weapon in the fight against cybercrime. An understanding of how various gangs operate and what tools they use helps build competent defenses and investigate incidents. This report takes a close look at the history of the Cuba group, and their attack tactics, techniques and procedures. We hope this article will help you to stay one step ahead of threats like this one.

Cuba ransomware gang

The group’s offensives first got on our radar in late 2020. Back then, the cybercriminals had not yet adopted the moniker “Cuba”; they were known as “Tropical Scorpius”.

Cuba mostly targets organizations in the United States, Canada and Europe. The gang has scored a series of resonant attacks on oil companies, financial services, government agencies and healthcare providers.

As with most cyberextortionists lately, the Cuba gang encrypts victims’ files and demands a ransom in exchange for a decryption key. The gang infamously uses complex tactics and techniques to penetrate victim networks, such as exploitation of software vulnerabilities and social engineering. They have been known to use compromised remote desktop (RDP) connections for initial access.

The Cuba gang’s exact origins and the identities of its members are unknown, although some researchers believe it might be a successor to another ill-famed extortion gang, Babuk. The Cuba group, like many others of its kind, is a ransomware-as-a-service (RaaS) outfit, letting its partners use the ransomware and associated infrastructure in exchange for a share of any ransom they collect.

The group has changed names several times since its inception. We are currently aware of the following aliases it has used:

• ColdDraw
• Tropical Scorpius
• Fidel
• Cuba

This past February, we came across another name for the gang — “V Is Vendetta”, which deviated from the hackers’ favorite Cuban theme. This might have been a moniker used by a sub-group or affiliate.

There is an obvious connection with the Cuba gang: the newly discovered group’s website is hosted in the Cuba domain:

http[:]//test[.]cuba4ikm4jakjgmkezytyawtdgr2xymvy6nvzgw5cglswg3si76icnqd[.]onion/

Cuba remains active as at the time of writing this, and we keep hearing about new extortion victims.

Victimology

In this section, we used data consensually provided by our users and information about victims from open sources, such as other security vendors’ reports and the data leak site of the ransomware gang itself.

The group has attacked numerous companies around the world. Industry affiliation does not seem to be a factor: victims have included retailers, financial and logistical services, government agencies, manufacturers, and others. In terms of geography, most of the attacked companies have been located in the United States, but there have been victims in Canada, Europe, Asia and Australia.

Ransomware

The Cuba ransomware is a single file without additional libraries. Samples often have a forged compilation timestamp: those found in 2020 were stamped with June 4, 2020, and more recent ones, June 19th, 1992.

Cuba extortion model

Four extortion models exist today in terms of tools used for pressuring the victim.

• Single extortion: encrypting data and demanding a ransom just for decryption.
• Double extortion: besides encrypting, attackers steal sensitive information. They threaten to both withhold the encryption key and publish the stolen information online unless the victim pays up. This is the most popular model among ransomware gangs today.
• Triple extortion: adding a threat to expose the victim’s internal infrastructure to DDoS attacks. The model became widespread after the LockBit gang got DDoS’ed, possibly by a victim. After getting targeted, the hackers realized that DDoS was an effective pressure tool, something they stated openly, setting an example for others. To be fair, isolated cases of triple extortion predate the LockBit case.
• The fourth model is the least common one, as it implies maximum pressure and is thus more costly. It adds spreading news of the breach among the victim’s investors, shareholders and customers. DDoS attacks in that case are not necessary. This model is exemplified by the recent hack of Bluefield University in Virginia, where the AvosLocker ransomware gang hijacked the school’s emergency broadcast system to send students and staff SMS texts and email alerts that their personal data had been stolen. The hackers urged not to trust the school’s management, who they said were concealing the true scale of the breach, and to make the situation public knowledge as soon as possible.

The Cuba group is using the classic double extortion model, encrypting data with the Xsalsa20 symmetric algorithm, and the encryption key, with the RSA-2048 asymmetric algorithm. This is known as hybrid encryption, a cryptographically secure method that prevents decryption without the key.

Cuba ransomware samples avoid encrypting files with the following name extensions: .exe, .dll, .sys, .ini, .lnk, .vbm and .cuba, and the following folders:

• \windows\
• \program files\microsoft office\
• \program files (x86)\microsoft office\
• \program files\avs\
• \program files (x86)\avs\
• \$recycle.bin\
• \boot\
• \recovery\
• \system volume information\
• \msocache\
• \users\all users\
• \users\default user\
• \users\default\
• \temp\
• \inetcache\
• \google\

The ransomware saves time by searching for, and encrypting, Microsoft Office documents, images, archives and others in the %AppData%\Microsoft\Windows\Recent\ directory, rather than all files on the device. It also terminates all SQL services to encrypt any available databases. It looks for data both locally and inside network shares.

Besides encrypting, the group steals sensitive data that it discovers inside the victim’s organization. The type of data that the hackers are after depends on the industry that the target company is active in, but in most cases, they exfiltrate the following:

• Financial documents
• Bank statements
• Company accounts details
• Source code, if the company is a software developer

Arsenal

The group employs both well-known, “classic” credential access tools, such as mimikatz, and self-written applications. It exploits vulnerabilities in software used by the victim companies: mostly known issues, such as the combination of ProxyShell and ProxyLogon for attacking Exchange servers, and security holes in the Veeam data backup and recovery service.

• Bughatch
• Burntcigar
• Cobeacon
• Hancitor (Chanitor)
• Termite
• SystemBC
• Veeamp
• Wedgecut
• RomCOM RAT

• Mimikatz
• PowerShell
• PsExec
• Remote Desktop Protocol

ProxyShell:

• CVE-2021-31207
• CVE-2021-34473
• CVE-2021-34523

ProxyLogon:

• CVE-2021-26855
• CVE-2021-26857
• CVE-2021-26858
• CVE-2021-27065

Veeam vulnerabilities:

• CVE-2022-26501
• CVE-2022-26504
• CVE-2022-26500

ZeroLogon:

• CVE-2020-1472

Profits

The incoming and outgoing payments in the bitcoin wallets whose identifiers the hackers provide in their ransom notes exceed a total of 3,600 BTC, or more than $103,000,000 converted at the rate of $28,624 for 1 BTC. The gang owns numerous wallets, constantly transferring funds between these, and uses bitcoin mixers: services that send bitcoins through a series of anonymous transactions to make the origin of the funds harder to trace.

Investigation of a Cuba-related incident and analysis of the malware

Host: SRV_STORAGE

On December 19, we spotted suspicious activity on a customer host, which we will refer to as “SRV_STORAGE” in this report. Telemetry data showed three suspicious new files:

An analysis of kk65.bat suggested that it served as a stager that initiated all further activity by starting rundll32 and loading the komar65 library into it, which runs the callback function DLLGetClassObjectGuid.

Let us take a look inside the suspicious DLL.

Bughatch

The komar65.dll library is also known as “Bughatch”, a name it was given in a report by Mandiant.

The first thing that caught our attention was the path to the PDB file. There’s a folder named “mosquito” in it, which translates into Russian as “komar”. The latter is a part of the DDL name suggesting the gang may include Russian speakers.

The DLL code presents Mozilla/4.0 as the user agent when connecting to the following two addresses:

• com, apparently used for checking external connectivity
• The gang’s command-and-control center. The malware will try calling home if the initial ping goes through.

This is the kind of activity we observed on the infected host. After Bughatch successfully established a connection with the C2 server, it began collecting data on network resources.

Looking into the C2 servers, we found that in addition to Bughatch, these spread modules that extend the malware’s functionality. One of those collects information from the infected system and sends it back to the server in the form of an HTTP POST request.

One could think of Bughatch as a backdoor of sorts, deployed inside the process memory and executing a shellcode block within the space it was allocated with the help of Windows APIs (VirtualAlloc, CreateThread, WaitForSingleObject), to then connect to the C2 and await further instructions. In particular, the C2 may send a command to download further malware, such as Cobalt Strike Beacon, Metasploit, or further Bughatch modules.

SRV_Service host

Veeamp

After some time, we found a malicious process started on a neighboring host; we dubbed this “SRV_Service”:

Veeamp.exe is a custom-built data dumper written in C#, which leverages security flaws in the Veeam backup and recovery service to connect to the VeeamBackup SQL database and grab account credentials.

Veeamp exploits the following Veeam vulnerabilities: CVE-2022-26500, CVE-2022-26501, CVE-2022-26504. The first two allow an unauthenticated user to remotely execute arbitrary code, and the third one, lets domain users do the same. After any of the three are exploited, the malware outputs the following in the control panel:

• User name
• Encrypted password
• Decrypted password
• User description in the Credentials table of Veeam: group membership, permissions and so on

The malware is not exclusive to the Cuba gang. We spotted it also in attacks by other groups, such as Conti and Yanluowang.

Activity we saw on SRV_Service after Veeamp finished its job was similar to what we had observed on SRV_STORAGE with Bughatch:

As was the case with SRV_STORAGE, the malware dropped three files into the temp folder, and then executed these in the same order, connecting to the same addresses.

Avast Anti-Rootkit driver

After Bughatch successfully established a connection to its C2, we watched as the group used an increasingly popular technique: Bring Your Own Vulnerable Driver (BYOVD).

The malicious actors install the vulnerable driver in the system and subsequently use it to various ends, such as terminating processes or evading defenses through privilege escalation to kernel level.

Hackers are drawn to vulnerable drivers because they all run in kernel mode, with a high level of system access. Besides, a legitimate driver with a digital signature will not raise any red flags with security systems, helping the attackers to stay undetected for longer.

During the attack, the malware created three files in the temp folder:

• aswarpot.sys: a legitimate anti-rootkit driver by Avast that has two vulnerabilities: CVE-2022-26522 and CVE-2022-26523, which allow a user with limited permissions to run code at kernel level.
• KK.exe: malware known as Burntcigar. The file we found was a new variety that used the flawed driver to terminate processes.
• av.bat batch script: a stager that helps the kernel service to run the Avast driver and executes Burntcigar.

Analysis of the BAT file and telemetry data suggests that av.bat uses the sc.exe utility to create a service named “aswSP_ArPot2”, specifying the path to the driver in the С\windows\temp\ directory and the service type as kernel service. The BAT file then starts the service with the help of the same sc.exe utility and runs KK.exe, which connects to the vulnerable driver.

Burntcigar

The first thing we noticed while looking into Burntcigar was the path to the PDB file, which contained a folder curiously named “Musor” (the Russian for “trash”), more indication that the members of the Cuba gang may speak Russian.

We further discovered that the sample at hand was a new version of Burntcigar, undetectable by security systems at the time of the incident. The hackers had apparently updated the malware, as in the wake of previous attacks, many vendors were able to easily detect the logic run by older versions.

You may have noticed that in the screenshot of our sample below, all data about processes to be terminated is encrypted, whereas older versions openly displayed the names of all processes that the attackers wanted stopped.

The malware searches for process names that suggest a relation to popular AV or EDR products and adds their process IDs to the stack to terminate later.

Burntcigar uses the DeviceIoContol function to access the vulnerable Avast driver, specifying the location of the code that contains the security issue as an execution option. The piece of code contains the ZwTerminateProcess function, which the attackers use for terminating processes.

Fortunately, our product’s self-defense was able to cope with the malware by blocking all hooks to the driver.

Later, we discovered similar activity exploiting the Avast anti-rootkit driver on the Exchange server and the SRV_STORAGE host. In both cases, the attackers used a BAT file to install the insecure driver and then start Burntcigar.

SRV_MAIL host (Exchange server)

On December 20, the customer granted our request to add the Exchange server to the scope of monitoring. The host must have been used as an entry point to the customer network, as the server was missing critical updates, and it was susceptible to most of the group’s initial access vectors. In particular, SRV_MAIL had the ProxyLogon, ProxyShell and Zerologon vulnerabilities still unremediated. This is why we believe that the attackers penetrated the customer network through the Exchange server.

On SRV_MAIL, the SqlDbAdmin user showed the same kind of activity as that which we had observed on the previous hosts.

We found that the attackers were using the legitimate gotoassistui.exe tool for transferring malicious files between the infected hosts.

GoToAssist is an RDP support utility often used by technical support teams, but the application is often abused to bypass any security defenses or response teams when moving files between systems.

We also found that new Bughatch samples were being executed. These used slightly different file names, callback functions and C2 servers, as our systems were successfully blocking older versions of the malware at that time.

SqlDbAdmin

We wondered who that SqlDbAdmin was. The answer came through a suspicious DLL, addp.dll, which we found manually on a compromised host.

We found that it used the WIN API function NetUserAdd to create the user. The name and password were hard-coded inside the DLL.

As we looked further into the library, we found that it used the RegCreateKey function to enable RDP sessions for the newly created user by modifying a registry setting. The library then added the user to the Special Account registry tree to hide it from the system login screen, an interesting and fairly unconventional persistence technique. In most cases, bad actors add new users with the help of scripts thatsecurity products rarely miss.

Cobalt Strike

We found a suspicious DLL, ion.dll, running on the Exchange server as part of the rundll32 process with unusual execution options. At first, we figured that the activity was similar to what we had earlier seen with Bughatch. However, further analysis showed that the library was, in fact, a Cobalt Strike Beacon.

When we were looking at the ion.dll code, what caught our attention was execution settings and a function that uses the Cobalt Strike configuration. The library used the VirtualAlloc function for allocating process memory to execute the Cobalt Strike Beacon payload in, later.

All configuration data was encrypted, but we did find the function used for decrypting that. To find the Cobalt Strike C2 server, we inspected a rundll32 memory dump with ion.dll loaded into it, running with the same settings it did on the victim host.

Finding out the name of the C2 helped us to locate the history of communications with that server within the telemetry data. After the malware connected to the C2, it downloaded two suspicious files into the Windows folder on the infected server and then executed these. Unfortunately, we were not able to obtain the two files for analysis, as the hackers had failed to disable security at the previous step, and the files were wiped off the infected host. We do believe, though, that what we were dealing with was the ransomware itself.

The customer promptly isolated the affected hosts and forwarded the incident to the Kaspersky Incident Response team for further investigation and search for possible artifacts. This was the last we saw of the malicious actor’s activity in the customer system. The hosts avoided encryption thanks to the customer following our recommendations and directions, and responding to the incident in time.

New malware

We found that VirusTotal contained new samples of the Cuba malware with the same file metadata as the ones in the incident described above. Some of those samples had successfully evaded detection by all cybersecurity vendors. We ran our analysis on each of the samples. As you can see from the screenshot below, these are new versions of Burntcigar using encrypted data for anti-malware evasion. We have made Yara rules that detect these new samples, and we are providing these in the attachment to this article.

BYOVD (Bring Your Own Vulnerable Driver)

We will now take a closer look at an attack that uses insecure drivers, which we observed as we investigated the incident and which is currently growing in popularity as various APT and ransomware gangs add it to their arsenals.

Bring Your Own Vulnerable Driver (BYOVD) is a type of attack where the bad actor uses legitimate signed drivers that are known to contain a security hole to execute malicious actions inside the system. If successful, the attacker will be able to exploit the vulnerabilities in the driver code to run any malicious actions at kernel level!

Understanding why this is one of the most dangerous kinds of attacks takes a quick refresher on what drivers are. A driver is a type of software that acts as an intermediary between the operating system and the device. The driver converts OS instructions into commands that the device can interpret and execute. A further use of drivers is supporting applications or features that the operating system originally lacks. As you can see from the image below, the driver is a layer of sorts between user mode and kernel mode.

Applications running in user mode have fewer privileges to control the system. All they can get access to is a virtualized memory area that is isolated and protected from the rest of the system. The driver runs inside the kernel memory, and it can execute any operations just like the kernel itself. The driver can get access to critical security structures and modify those. Modifications like that make the system liable to attacks that use privilege escalation, disabling of OS security services, and arbitrary reading and writing.

The Lazarus gang made use of that technique in 2021 as they gained write access to kernel memory and disabled Windows security features by abusing a Dell driver that contained the CVE-2021-21551 vulnerability.

There is no sure-fire defense from legitimate drivers, because any driver could prove to have a security flaw. Microsoft has published a list of recommendations to protect against this type of techniques:

• Enable Hypervisor-Protected Code Integrity.
• Enable Memory Integrity.
• Enable validation of driver digital signatures.
• Use the vulnerable driver blocklist.

However, studies suggest that the recommendations are irrelevant even with every Windows protection feature enabled, and attacks like these go through anyway.

To counter this technique, many security vendors started adding a self-defense module into their products that prevents malware from terminating processes and blocks every attempt at exploiting vulnerable drivers. Our products have that feature too, and it proved effective during the incident.

Conclusion

The Cuba cybercrime gang employs an extensive arsenal of both publicly available and custom-made tools, which it keeps up to date, and various techniques and methods including fairly dangerous ones, such as BYOVD. Combating attacks at this level of complexity calls for sophisticated technology capable of detecting advanced threats and protecting security features from being disabled, and a massive, continuously updated threat knowledge base that helps to detect malicious artifacts manually.

The incident detailed in this article shows that investigation of real-life cyberattacks and incident response, such as Managed Detection and Response (MDR), are sources of the latest information about malicious tactics, techniques and procedures. In particular, during this investigation, we discovered new and previously undetected samples of the Cuba malware, and artifacts suggesting that at least some of the gang members spoke Russian.

That said, effective investigation and response begin with knowledge of current cyberthreats, which is available from Threat Intelligence services. At Kaspersky, the Threat Intelligence and MDR teams work closely while exchanging data and enhancing their services all the time.

Appendix

Sigma and YARA rules: https://github.com/BlureL/SigmaYara-Rules
Indicators of Compromise: Download PDF
Mitre ATT&CK matrices: Download PDF