0 %

A lookback under the TA410 umbrella: Its cyberespionage TTPs and activity

April 28, 2022

ESET researchers reveal an in depth profile of TA410: we imagine this cyberespionage umbrella group consists of three totally different groups utilizing totally different toolsets, together with a brand new model of the FlowCloud espionage backdoor found by ESET.

ESET researchers have documented and analyzed TA410 exercise going again to 2019. TA410 is a cyberespionage umbrella group loosely linked to APT10, recognized largely for concentrating on US-based organizations within the utilities sector, and diplomatic organizations within the Center East and Africa. TA410 has been energetic since a minimum of 2018 and was first publicly revealed in August 2019 by Proofpoint in its LookBack blogpost. A 12 months later, the then-new and really complicated malware household known as FlowCloud was additionally attributed to TA410.

On this blogpost, we offer an in depth profile of this APT group, together with its modus operandi and toolset that features a new model of FlowCloud, found by ESET. This very complicated backdoor comprises fascinating espionage capabilities. ESET will current its newest findings about TA410, together with outcomes from ongoing analysis, throughout Botconf 2022. For YARA and Snort guidelines, seek the advice of ESET’s GitHub account.

Key factors on this blogpost:

  • TA410 is an umbrella group comprised of three groups ESET researchers named FlowingFrog, LookingFrog and JollyFrog, every with its personal toolset and targets.
  • ESET telemetry exhibits victims all world wide, primarily within the governmental and schooling sectors.
  • TA410 had entry to the newest recognized Microsoft Trade distant code execution vulnerabilities, e.g., ProxyLogon in March 2021 and ProxyShell in August 2021.
  • ESET researchers discovered a brand new model of FlowCloud, a posh and modular C++ RAT. It has a number of fascinating capabilities, together with:
    • Controlling related microphones and triggering recording when sound ranges above a specified threshold quantity are detected.
    • Monitoring clipboard occasions to steal clipboard content material.
    • Monitoring file system occasions to gather new and modified recordsdata.
    • Controlling hooked up digital camera units to take photos of the compromised pc’s environment.
  • FlowCloud deploys a rootkit to cover its exercise on the compromised machine.
  • The LookBack backdoor utilized by TA410 makes use of a customized community protocol, which might perform over HTTP or uncooked TCP, for C&C server communications.
  • TA410 is likely one of the customers of the Royal Street malicious doc builder.

TA410 groups compromise their targets in varied methods, which signifies to us that these victims are focused particularly, with the attackers selecting which entry methodology has the very best probability of infiltrating the goal.

The primary stage of the FlowCloud model recognized by ESET researchers can examine whether or not particular safety software program is put in on the machine it tries to compromise, however this isn’t applied within the loaders we analyzed. Nonetheless, we discovered a customized AntivirusCheck class, which might examine working processes in opposition to a hardcoded record of executable filenames from recognized safety merchandise, together with ESET merchandise. In case certainly one of these merchandise is detected, FlowCloud goes by means of its common loading course of and cancels the auto_start_after_install configuration worth.

Regardless that we imagine that this model of FlowCloud continues to be present process growth and testing, the cyberespionage capabilities of this model embrace the flexibility to gather mouse actions, keyboard exercise, and clipboard content material together with details about the present foreground window. This info can assist attackers perceive stolen knowledge by contextualizing it.

FlowCloud may also collect details about issues occurring across the sufferer’s pc by taking photos utilizing related digital camera peripherals and recording audio utilizing a pc’s microphone. This latter perform is triggered by any sound over a threshold of 65 decibels, which is within the higher vary of regular dialog quantity.


ESET researchers imagine that TA410 consists of three totally different groups, utilizing very related techniques, methods, and procedures (TTPs) however totally different toolsets and exiting from IP addresses positioned in three totally different districts. These groups, referred to under as FlowingFrog, LookingFrog, and JollyFrog, have overlaps in TTPs, victimology and community infrastructure.

  • FlowingFrog makes use of Royal Road RTF documents, a first-stage implant known as Tendyron, and a really complicated second-stage backdoor known as FlowCloud.
  • LookingFrog makes use of a first-stage backdoor known as X4, and LookBack as a second stage.
  • JollyFrog makes use of solely generic malware households corresponding to Korplug (aka PlugX) and QuasarRAT. A part of the exercise of this staff was described by Fortinet, who attributed the exercise to APT10. ESET researchers, nonetheless, imagine this exercise is totally different from the operations that APT10 (aka A41APT) has carried out just lately.

FlowingFrog and JollyFrog share community infrastructure – extra exactly, the area ffca.caibi379[.]com, as talked about by Proofpoint.

FlowingFrog and LookingFrog ran a phishing marketing campaign on the identical time in opposition to the identical targets, as additionally talked about in the identical Proofpoint article.

In ESET telemetry, we don’t see some other overlap between these subgroups. We imagine that these subgroups function considerably independently however that they could share intelligence necessities, an entry staff that runs their spearphishing campaigns, and likewise the staff that deploys community infrastructure.


Most TA410 targets are high-profile organizations within the diplomacy and schooling sectors, however we’ve got additionally seen victims within the navy sector, a producing firm in Japan, a mining firm in India, and a charity in Israel. In accordance with ESET telemetry, the victims are positioned in Africa, Asia, the Center East, and Europe. Apparently, there is no such thing as a clear segmentation of the concentrating on (by sector or geography) among the many totally different groups.

A component price mentioning is that TA410 targets international people in China. In ESET telemetry, we’ve got noticed this as having occurred a minimum of twice: as an illustration, one sufferer is a French educational, and one other is a member of a diplomatic mission of a South Asian nation in China.

Since 2018, we’ve got seen the next targets, additionally depicted in Determine 1:

  • FlowingFrog: College, international diplomatic mission of a South Asian nation in China, mining firm
  • LookingFrog: Diplomatic missions, charity, authorities and industrial manufacturing
  • JollyFrog: Training, church, navy, diplomatic mission

Determine 1. Map of nations and verticals focused by TA410

Preliminary compromise and typical TTPs

If we exclude the totally different backdoors, the three groups use the same modus operandi. They compromise their targets both by spearphishing, in response to Proofpoint, or, for LookingFrog and JollyFrog, by compromising a web-facing utility corresponding to Microsoft Trade or SharePoint. This might point out that victims are focused particularly, with the attackers selecting which entry methodology is the very best for a given goal.

The general public-facing utility compromise strategy is what we’ve got seen probably the most. Attackers linked to LookingFrog exploited Microsoft SharePoint servers in 2019 to achieve code execution, in all probability by leveraging CVE-2019-0604. They then dropped an ASPX webshell that was used to put in different malicious parts. These had been both dropped straight by way of the webshell or downloaded from a distant server utilizing certutil.exe, a recognized LOLBin.

In 2020, we noticed additional exploitations by JollyFrog, of Microsoft SQL servers and IIS servers working customized purposes.

In August 2021, we noticed LookBack being loaded by an IIS employee course of on a server belonging to an industrial manufacturing firm in Japan. This occurred following the exploitation of the Trade ProxyShell vulnerability on that server, as we describe in ESET Menace Report T3 2021.

This exhibits that LookingFrog operators carefully observe the invention of RCE vulnerabilities in widespread server purposes and rapidly make use of any obtainable exploit with the intention to achieve management of unpatched servers run by organizations on their goal lists.

Along with the full-featured backdoors analyzed within the following sections, these attackers use quite a lot of instruments corresponding to vulnerability scanners, exploits from the Equation Group leaks, proxy/tunneling utilities (HTran, LCX, EarthWorm), and lateral motion scripts corresponding to WMIExec.


TA410 – FlowingFrog

FlowingFrog makes use of a primary stage that ESET researchers have named the Tendyron downloader, and a posh second stage named FlowCloud, so named by the builders in its modules’ PDB paths.

Royal Street and Tendyron downloader

Royal Street is a malicious doc builder utilized by a number of cyberespionage teams (see the analysis by nao_sec). Recordsdata constructed with this instrument are RTF paperwork exploiting Equation Editor N-day vulnerabilities corresponding to CVE-2017-11882. TA410 operators at all times use the Royal Street encoding bytes: A9 A4 6E FE, as seen in Determine 2.

Determine 2. Encoded Royal Street payload

On October 13th 2020, we observed {that a} new Royal Street RTF doc, proven in Determine 3, had been uploaded to VirusTotal.

Determine 3. Royal Street RTF doc discovered on VirusTotal (SHA‑1: ADD5B4FD9AEA6A38B5A8941286BC9AA4FE23BD20)

When opened, the doc triggers the injection of a customized downloader – a PE executable – into an iexplore.exe course of. The PE sources 103, 104 and 105 comprise the payload URLs, XORed with 0xD3. The next recordsdata are downloaded and written to disk:

  • http://103.139.2[.]93:1702/tdr.dat written to %localappdatapercentTendyronTendyron.exe
    (SHA-1: 09C76522136B5E9BAB74381FEEE265F7E9B1D550)
  • http://103.139.2[.]93:1702/okt.dat written to %localappdatapercentTendyronOnKeyToken_KEB.dll (SHA‑1: F359D3C074135BBCA9A4C98A6B6544690EDAE93D)
  • http://103.139.2[.]93:1702/md.dat written to %localappdatapercentTendyronTendyron.conf
    (we weren’t capable of retrieve this file)

Lastly, this course of individually downloads http://103.139.2[.]93:1702/t86.dat (useful resource 101), hundreds it into reminiscence, and calls its startModule export. Sadly, we weren’t capable of retrieve this pattern.

Tendyron.exe is a reputable executable, signed by online-banking safety vendor Tendyron Company, and that’s weak to DLL search-order hijacking. Persistence for the downloaded payload is established by way of the Tendyron worth below the Run key HKCUSOFTWAREMicrosoftWindowsCurrentVersionRun.

When executed, Tendyron.exe hundreds the malicious OnKeyToken_KEB.dll. The export OnKeyT_ContextInit comprises code that decrypts hardcoded shellcode (see Determine 4) and injects it into iexplore.exe utilizing WriteProcessMemory.

Determine 4. Shellcode decryption loop

The following stage, injected into iexplore.exe, is a backdoor written utilizing the Microsoft Basis Class (MFC) framework. It additionally comprises RTTI symbols and thus just a few C++ class names:

  • ClientSocket
  • Supervisor
  • DllManager
  • KernelManager

These class names are the identical as utilized in Farfli/Gh0stRAT, a backdoor that has been used for greater than 10 years to conduct (largely) cyberespionage operations. Its supply code was leaked and is now obtainable on GitHub. Thus, we imagine that TA410 builders reused code copied from Farfli.

The C&C server is hardcoded, in cleartext, within the pattern; on this particular case, it’s set to 114.118.83[.]141.

On VirusTotal, as proven in Determine 5, we are able to see yet one more HTTP request to 103.139.2[.]93 was triggered in the course of the execution of the RTF file. The results of the request to http://103.139.2[.]93:1702/SL3716/S8437AEB.DAT was recorded by VirusTotal and the SHA-1 of this encrypted file is 140F81037A76B7B16A00E1D5E0E2CD9F6687F642. This URI is typical of these used to obtain FlowCloud, a posh C++ implant described within the subsequent part.

Determine 5. URL requests seen by the VirusTotal sandbox throughout execution of the malicious RTF doc

The equivalent encrypted file was additionally downloaded from http://114.55.109[.]199:56022/SL3716/S8437AEB.DAT by a FlowCloud dropper model 4.1.3 (SHA‑1: 014421BDB1EA105A6DF0C27FC114819FF3637704). A abstract of the compromise chain is supplied in Determine 6.

Determine 6. Compromise chain from the Royal Street doc to FlowCloud


FlowCloud is a posh implant written in C++. It consists of three fundamental parts, deployed in a multistage course of that makes use of varied obfuscation and encryption methods to hinder evaluation. A number of variations of FlowCloud have been recognized since 2020, most notably variations 4.1.3 and 5.0.1 described by Proofpoint. On this part, we analyze FlowCloud variations 5.0.2 and 5.0.3. Opposite to these beforehand discovered, the samples we obtained for model 5.0.2 comprise verbose error messages and meticulous logging.

This deployment course of is similar to the one described by Proofpoint for model 5.0.1. The three fundamental parts are a driver with rootkit performance, a easy persistence module, and a customized backdoor. We describe these intimately within the upcoming sections.

Loader (ClientLdrExe)

The primary stage is accountable largely for creating the recordsdata and registry keys utilized by the opposite phases. The values for these executables and configuration knowledge might be discovered, encrypted, within the loader’s useful resource part. Desk 1 comprises an summary of those sources and their use.

Desk 1. Contents of the dropper’s sources

Useful resource ID Position Inner title
100 FlowCloud RAT DLL fcClientDll 
101 32-bit rootkit driver Driver
102 64-bit rootkit driver Driver
103 DLL hijacking weak app N/A
104 Shellcode loaded by the malicious library within the DLL hijacking SETLANG_dlcore
105 Shellcode that hundreds fcClient (unused) N/A
106 Ultimate dropper stage fcClient
107 32-bit persistence module fcClientWD_x86
108 64-bit persistence module fcClientWD_x64
109 Reliable library used for module stomping slam
110 DLL used for hijacking XXXModule_dlcore0
1000 Protobuf serialized FlowCloud configuration N/A
1001 Dropper configuration N/A
2000 Used as a substitute or extension to useful resource 2001 N/A
2001 Path to the registry key for the PrintProcessor service (utilized by the driving force) N/A
10000 Set up configuration N/A

Within the situations we noticed, most sources are written to disk encrypted, and solely decrypted in reminiscence when wanted. In some circumstances, they’re then re-encrypted however with a unique key. This system makes it tougher to dump the plaintext values from the method’s reminiscence and to research exit dumps. The paths and registry keys to make use of, and whether or not they need to be decrypted earlier than being written, are outlined within the set up configuration. The samples we analyzed all retailer their recordsdata within the %ProgramFilespercentMSBuildMicrosoftExpressionBlendmsole listing; we imagine that that is the default worth. FlowCloud makes use of filenames which can be both much like these of reputable Home windows recordsdata (e.g., rebare.dll which may very well be mistaken for rebar.dll) or innocuous wanting (e.g., AC146142) to keep away from suspicion.

Determine 7 presents a graphical overview of the deployment course of and its parts. We clarify every of the steps in additional element within the upcoming sections.

Determine 7. FlowCloud deployment course of

First, the loader decrypts and parses the embedded set up configuration, which makes use of the Home windows INI format. This configuration defines the malware’s set up path together with the filename or registry key the place every embedded useful resource is to be written. The identical values are hardcoded within the following phases, which leads us to suppose that the samples are generated utilizing a builder. In a pattern we analyzed, this configuration is accompanied with feedback explaining the values for some sections. Determine 8 exhibits this set up configuration with the feedback translated to English.

Determine 8. Set up configuration with explanatory feedback. Observe that some fields are commented out.

The configuration may also comprise a bit defining particular safety software program to examine for, however this isn’t applied within the loaders we analyzed. Nonetheless, there’s a customized AntivirusCheck class, which might examine working processes in opposition to a hardcoded record of XOR-encrypted executable filenames from recognized safety merchandise: 360 Whole Safety, Avast, Avira, AVG, Bitdefender, ESET, Jiangmin Expertise Antivirus, Kingsoft, McAfee, Micropoint, Norton, Rising Antivirus, and Development Micro. This class is just used if the loader is about to straight begin the fcClient module by way of the auto_start_after_install configuration key.

Relying on the configuration keys used, the loader can both load the fcClientDll RAT module straight, thus bypassing many of the complicated deployment course of, or it could actually create a service or scheduled activity. Within the former case, the duty or service attains persistence by being set to begin robotically on boot. Within the samples we noticed, the duty or service was configured to execute the following step of the set up course of by working a reputable utility weak to DLL search-order hijacking. The appliance and the accompanying related and malicious DLL had been each embedded within the loader’s sources.

DLL side-loading (XXXModule_dlcore0)

Within the samples we analyzed, the weak utility was both setlang.exe from Microsoft Workplace 2003 with a malicious setlangloc.dll or vpreview.exe from Visio Preview 2007 with a malicious vviewres.dll. Strings contained within the malicious DLL additionally level to emedres.dll from Emurasoft’s EmEditor as a potential third goal for DLL side-loading. This can be a actual risk as such vulnerabilities had been current in older variations of EmEditor, however we didn’t see any samples utilizing it.

In all noticed samples, the malicious library is identical and serves to load and execute shellcode from a file that’s saved below the identical title because the DLL, however with a .dat extension. We analyze this shellcode within the subsequent part, however first, we need to have a look at the notable anti-analysis methods used on this library.

Regardless of its comparatively easy objectives, the library’s code makes heavy use of anti-debugging methods and management circulate obfuscation to hinder evaluation. Within the perform that hundreds the following file, the helpful code is repeatedly interspersed with the identical sequence of opcodes to obfuscate this system’s circulate. As proven in Determine 9, this brief snippet is packed stuffed with anti-analysis methods, however in the end quantities to an unconditional 16-byte soar. This is sufficient to foil many automated analyses, together with decompilers.

Determine 9. Annotated disassembly of the management circulate obfuscation snippet

The above snippet is bookended by calls to 2 anti-debugging capabilities, as might be seen in Determine 10. The perform, which we named crash_if_debugger within the earlier screenshot, calls IsDebuggerPresent and checks some generally hooked library capabilities for a breakpoint as their first instruction. If these checks detect a debugger, the perform returns a price that may trigger this system to leap to an invalid deal with and crash. The second raises an exception by way of the INT 0x2D instruction and exits if it was dealt with by a debugger.

Determine 10. Decompiler view displaying the plain sample of anti-debugging checks. Observe that we needed to take away the aforementioned obfuscation for the decompiler to provide any output.

fcClient (rescure.dat)

When it’s first executed, this module units up persistence and installs the backdoor, rootkit, and persistence modules. It then units particular registry keys and recordsdata as guardrails to skip the setup on subsequent runs.

First, persistence is established through the use of the ITaskService COM interface to create the MicrosoftWindowsCertificateServicesClientNetTask scheduled activity. If a activity with the identical title already exists, it’s deleted earlier than the brand new one is created. This activity will run the DLL hijacking goal as SYSTEM at every boot.

Afterwards, the rootkit module is decrypted and written to the %Systempercentdrivers folder as hidmouse.sys. A hidmouse service is then created to run that module and is straight away began. The file is then deleted from the disk and changed by a replica of the reputable hidusb.sys driver from the identical folder. Thus, anybody wanting on the file on disk moderately than the one mapped into reminiscence would see a reputable, benign file.

On Home windows 10 machines, the system time is briefly modified to make it appear like the service was created in January 2013. Each this and using the reputable driver listing assist the rootkit mix in with different drivers.

The next recordsdata are copied to the %System% listing:

  • The backdoor: rescure.dat
  • A decoy DLL: sspisrvui.dat as sspisrvui.dll (timestomped to July 2013)
  • The encrypted shellcode: rebare.dat

The rebare.dat shellcode is similar to that used within the self-decrypting DLL, however it hundreds fcClient straight.

FlowCloud then begins a suspended course of to carry out injection on it. This course of is created by way of CreateProcessAsUserW utilizing a token retrieved from the explorer.exe or winlogon.exe course of within the present session.

The injected code hundreds the identical backdoor (rescure.dat) into the method’s reminiscence and calls its startModule export to complete the set up. In the meantime, the injection course of is terminated.

At this level, set up of the backdoor is full. All that’s left is to execute the backdoor. To attain this, the brand new course of hundreds the decoy DLL and manually replaces its content material in reminiscence with the fcClientDll module (a course of often known as module stomping or DLL hollowing), earlier than calling its fundamental perform.

fcClientDll (responsor.dat)

This complicated module is the principle element of the backdoor. It offers a variety of capabilities from full file system entry to manage of digital camera peripherals and the whole lot in between. Though we didn’t observe any plug-ins, the backdoor comprises code that hints that they can be utilized to additional lengthen performance.

Earlier than diving deeper into the functionalities, we need to spotlight some notable traits:

  • Configuration info and knowledge for communications with the C&C server are Protobuf-serialized, compressed, and encrypted.
  • File exfiltration is finished by means of encrypted, Protobuf-serialized constructions and is disguised as HTTP by prepending the information with a hardcoded, faux POST request. The Content material-Size header is the one variable component, as it’s set to the precise measurement of the information despatched. This hardcoded request might be seen in Determine 11.
  • A number of functionalities are applied by means of using COM objects and interfaces.

Determine 11. Hardcoded, faux HTTP POST request used for FlowCloud C&C communication

This element makes use of an encrypted, Protobuf-serialized configuration that it tries to learn from a file on disk or a registry key. The configurations we noticed had been composed of three sections:

  1. server_config: This part comprises details about the C&C servers and identification details about the sufferer and backdoor.
  2. policys [sic]: This part defines the conduct of the backdoor’s parts and is described intimately within the following paragraphs.
  3. install_config: Because the title signifies, this part defines the set up parameters.

An instance of such a server_config is proven in Determine 12. This configuration corresponds to useful resource 1000 within the preliminary loader. It defines the deal with and port for each the exfiltration server (file_server) and the C&C server (exchange_server), together with the encryption key to make use of for communication with every. A fallback server can be outlined for every of those. The file_key discipline defines the encryption key to make use of when storing recordsdata which can be to be exfiltrated. The opposite entries are used to determine the backdoor and the victimized host:

  • product_name: A reputation for the backdoor in use. PCArrowI appears to correspond to FlowCloud.
  • product_version: The backdoor’s model.
  • id_prefix: This worth is prefixed to the generated ID. Presumably, used to group victims or campaigns.
  • id: This worth uniquely identifies the sufferer. Initially, it’s empty; the worth is generated on the primary execution utilizing the next format: __

Determine 12. server_config part of a decoded FlowCloud configuration

FlowCloud’s capabilities are unfold out over a collection of singleton courses, every of which implements a cohesive set of functionalities associated to a selected kind of knowledge or motion. These roughly observe an inner naming conference the place courses with names ending with manager_handler carry out actions in response to C&C instructions, whereas these whose names finish with supervisor robotically carry out actions based mostly on timers or occasion listeners.

Every supervisor shops collected knowledge in its personal SQLite database, whereas knowledge that’s collected on demand is returned on to the C&C server. Information is encrypted with the aforementioned file_key earlier than being inserted into the database. The situation of the SQLite databases is outlined by the data_folder set up configuration key, with the default worth being %ProgramFilespercentMSBuildMicrosoftExpressionBlendmsolefcdata.

The courses are orchestrated by an occasion of fc_kernel_manager. This object is accountable for initializing different parts and dealing with C&C connections. It could actually additionally replace the native configuration when the corresponding command is obtained.

As proven in Determine 13, parameters and frequency of automated actions might be specified and finely tuned by means of configuration insurance policies. Information exfiltration is likewise automated: insurance policies can comprise a cache_size or cache_count parameter, which determines how a lot knowledge might be collected regionally by the corresponding class earlier than it’s staged for exfiltration.

Determine 13. The policys [sic] part of a decoded FlowCloud config

As we’ve got beforehand talked about, this implant makes use of numerous courses. Fairly than documenting every of them individually, we’ll current an summary of the obtainable performance by grouping them into three classes: people who work together with the file system, functionalities that gather details about packages and processes, and people who collect real-time details about person exercise.

File system

FlowCloud offers interplay with the file system in quite a lot of methods, most of which might retailer file metadata and content material of their SQLite database.

Certainly one of these is a element that walks by means of all mapped file methods and collects recordsdata that aren’t excluded by filters within the smfile_search_policy. It additionally creates an invisible window that listens for file creation, modification, or renaming occasions. The corresponding recordsdata are collected until they’re excluded by that coverage.

One other element collects details about mapped volumes, together with mount level, title, drive kind, and disk utilization knowledge. This identical class collects file and listing metadata.

As a complement to those automated measures, the backdoor implements capabilities that present full entry and management over the content material of mounted drives. This consists of bidirectional file transfers between the C&C and the compromised machine.

Applications and processes

FlowCloud is ready to robotically receive a listing of put in software program by means of using the undocumented IShellAppManager COM interface. This performance can be invoked by way of a C&C command. Determine 14 exhibits, after the extraneous code has been eliminated, how that interface is used.

Determine 14. Simplified code displaying how the IShellAppManager COM interface is used to record put in purposes

Different instructions can be utilized to retrieve an in depth record of accessible providers and presently working processes.

One other fascinating function is the close to real-time monitoring of course of exercise. To attain this, FlowCloud runs WMI queries each second to get all course of creation and termination occasions. The obtained info is correlated with knowledge from the Win32_Process desk for a extra detailed view.

Consumer exercise

FlowCloud is in a position gather a miscellany of knowledge that we’ve got determined to group below the “Consumer exercise” umbrella.

It has the flexibility to watch the clipboard for modifications and save any knowledge it comprises. As seen in Determine 15, it achieves this by creating an invisible window with a customized class and registering two clipboard codecs. This window makes use of AddClipboardFormatListener (on Home windows Vista or more moderen) or SetClipboardViewer (on Home windows XP and prior) to pay attention for clipboard content material modifications.

Determine 15. Arrange monitoring of the clipboard

Collected clipboard content material is saved together with details about the present foreground window. This info can assist attackers perceive the information by contextualizing it.

FlowCloud can periodically take screenshots and retailer them with details about the foreground course of and time for the reason that final person enter. To restrict the disk area used, photos the place fewer than 5% of the pixels differ from probably the most just lately saved seize aren’t saved. This function can be invoked on demand by the server.

One other of the backdoor’s parts information mouse and keyboard exercise to a database. It doesn’t gather these straight, however as an alternative acts in tandem with the keylogger element of the driving force (described within the subsequent part) by studying knowledge from the .pipenamedpipe_keymousespy_english named pipe.

Apparently, FlowCloud may also collect details about issues occurring across the sufferer’s pc. The primary manner it does so is thru a C&C command that takes an image utilizing related digital camera peripherals. This function is applied utilizing the CCameraDS class from OpenCV.

The second manner it could actually gather details about the pc’s environment is by recording audio. Very similar to a voice assistant, FlowCloud can use a pc’s microphone to take heed to its environment, however as an alternative of recording being triggered by a command phrase, it appears to be triggered by any sound over a threshold outlined by the decibel_limit discipline of the audio_policy. The default worth is 65 decibels, which is within the higher vary of regular dialog quantity (commonly defined to be anyplace between 50 and 70 dB by varied sources).

Self-decrypting DLL (setlangloc.dat)

The loaded shellcode is a self-decrypting DLL. It first decrypts the embedded DLL utilizing a byte-oriented XOR-and-ADD scheme (proven in Determine 16). The shellcode we analyzed used the important thing 0x7B. As soon as it has decrypted the embedded DLL, the shellcode manually performs the capabilities of LoadLibrary and calls the loaded module’s startModule export.

Determine 16. Pseudocode for the DLL decryption routine

This newly loaded module makes use of the identical anti-debugging and anti-analysis methods because the hijacking DLL described above. On prime of these, it additionally makes use of just a few methods of its personal:

  • Covers its tracks by overwriting the code beforehand modified by the malicious library with a ineffective name to lstrlenW.
  • Base64-encoded strings are used for perform imports (by way of GetProcAddress) and solely decoded as wanted.
  • Exits if the method’s executable shouldn’t be the anticipated DLL hijacking goal (e.g., setlang.exe).

The module creates a brand new course of utilizing the identical executable and performs course of injection on it, redirecting the present thread to the written code area. This code inside the brand new course of launches a thread that decrypts and hundreds the fcClient module earlier than calling its startModule export. That perform will carry out the ultimate phases of the set up and cargo the DLL containing the backdoor performance.

Driver (hidmouse.sys)

FlowCloud’s driver serves a twin objective: it acts as each a keylogger and a rootkit. It accomplishes this primarily by hijacking native drivers’ handler capabilities for particular I/O management codes and changing them with its personal:

  • Learn (IRP_MJ_READ) for the keyboard driver (kbdclass or KeyboardClass0)
  • Learn (IRP_MJ_READ) for the mouse driver (mouclass or PointerClass0)
  • Machine management (IRP_MJ_DEVICE_CONTROL) for the community driver (tcpip or nsiproxy)

The motive force additionally offers kernel-level functionalities for use by the RAT. They are often invoked by way of IO management codes or by writing to particular registry keys.

This module is signed with a certificates with the thumbprint 02ED6A578C575C8D9C72398E790354B095BB07BC. Issued to Hangzhou Leishite Laser Expertise Co. in 2012 by Wosign and revoked in 2014, it appears most definitely this certificates was stolen.


In its IRP_MJ_READ handlers for keyboard and mouse occasions, the driving force merely information IO occasions to lookaside lists earlier than passing them to the reputable handler. This ensures that the driving force doesn’t intrude in a manner that may very well be noticeable by the person. These occasions are then parsed to the format utilized by the backdoor’s keymouse_manager and written to the named pipe .pipenamedpipe_keymousespy_english.


After hijacking the aforementioned drivers, the rootkit erases the DLL names related to them from inner constructions used to show gadget drivers.

The rootkit can stop processes from being proven by utilities that record working processes, corresponding to Process Supervisor. As proven in Determine 17, it achieves this by eradicating their entries from the ActiveProcessLinks record of the undocumented KPROCESS kernel construction. Since this construction shouldn’t be a part of the general public API and may change between releases, the rootkit comprises code to match the working system’s construct quantity to the proper offsets on this construction. That code covers all variations from Home windows XP to Home windows 10 20H1. This performance might be invoked on any course of by way of the IOCTL_HIDE_PROCESS_BY_PROCESSID (0x222028) management code. It is usually used, on driver startup, to cover the method with the PID contained within the registry key HKLMHARDWARE{76BA14B7-AF0C-4dc9-9E9D-2A6970F231D9}. This course of is additional camouflaged by altering its related executable filename to certainly one of svchost.exe or dllhost.exe in the identical kernel construction.

Determine 17. Perform used to forestall a course of from being displayed in lists of working processes

By way of its hijacking of the community driver, the rootkit may also disguise a single course of’s community site visitors from native utilities. The method whose site visitors is to be hidden is about by means of the IOCTL_SET_TRAFFICHIDE_PROCESSID (0x222048) management code.

Among the rootkit’s capabilities are utilized by the fcClientDll module to cover the method wherein it’s working.

Management codes to govern a course of title in varied inner constructions are additionally uncovered by the driving force.

Persistence module (fcClientWD)

This module is comparatively easy in comparison with different parts. The beforehand talked about NetTask already accomplishes persistence normally, by executing on system startup. This module enhances that mechanism by making certain persistence in a really particular edge case the place execution of the malware is perhaps interrupted: the person logs out on a system with hibernation and Fastboot enabled. On methods the place both of these is disabled, this module does nothing.

FlowCloud v4.1.3

This older model of FlowCloud has already been described in a Proofpoint blogpost and presents similarities to the newer model described within the previous subsections, so we’ll solely spotlight notable variations and new info revealed by our evaluation.

This model runs a number of anti-analysis and anti-detection checks earlier than executing its payload, and terminates if any of these assessments detect that the method is being analyzed. It checks working processes for executables of a number of recognized cybersecurity distributors. Whereas most of those names are additionally current in model 5, this record shouldn’t be a strict subset of the one v.5 makes use of. This tends to help the proposition that variations 4 and 5 of FlowCloud are maintained in parallel.

It additionally embeds a DLL model of the Pafish (aka Paranoid Fish) sandbox and evaluation detection instrument as certainly one of its encrypted sources. This library is loaded in reminiscence and the entire anti-analysis/anti-sandboxing checks it implements are run.

Apparently, the driving force put in is identical because the one for model 5.0.2. These utilized by model 5.0.3 present equivalent performance, however differ barely.

TA410 – LookingFrog

LookingFrog makes use of two fundamental malware households: X4 and LookBack. We have now seen each of them on machines belonging to the identical sufferer.


X4 is a customized backdoor that’s used as a primary stage, earlier than LookBack is deployed. It’s loaded by a VMProtect-ed loader, often named PortableDeviceApi.dll or WptsExtensions.dll. Sadly, we weren’t capable of uncover any persistence methodology.

The loader injects an orchestrator into reminiscence in a svchost.exe course of. In flip, the orchestrator injects the community element into reminiscence and communicates with it by way of a file positioned at C:ProgramDataMicrosoftCryptoRSAMachineKeysLogrsa.txt. Determine 18 exhibits a abstract of the X4 parts.

Determine 18. Abstract of the X4 parts

The community element is shellcode. It’s encrypted utilizing the AES algorithm and saved within the Home windows registry. Desk 2 exhibits the three registry keys utilized by X4.

Desk 2. Community shellcode registry keys

Registry Key Description
HKLMSOFTWAREMicrosoftDRMPSKey Title of the method into which the shellcode will probably be injected (spoolsv.exe).
HKLMSOFTWAREMicrosoftDRMX4Data Encrypted shellcode.

The decrypted shellcode appears prefer it was based mostly on Metasploit and communicates with a hardcoded IP deal with by way of HTTP. An fascinating attribute is that it makes use of the faux Host header onedrive.reside.com.

Each second, the orchestrator, which lives in reminiscence solely, reads the cleartext rsa.txt file to examine whether or not there are new instructions to execute. The instructions are obtained from the C&C server, by way of the community shellcode. Within the orchestrator, the instructions are recognized by a numerical identifier that’s computed from the command title, as proven in Determine 19.

Determine 19. Customized hash perform seen in X4

The orchestrator handles seven instructions, detailed in Desk 3. Output of those instructions is written to C:ProgramDataMicrosoftCryptoRSAMachineKeysLogoutput.log.

Desk 3. X4 backdoor instructions

ID Title Description
0x3ECFF9B9D92 osload Write new encrypted shellcode to HKLMSOFTWAREMicrosoftDRMX4Data. It could actually additionally modify X4Key and PSKey.
0x3F5FAFC0EDD pskill Kill a course of by PID.
0x3F5FB1E6015 pslist Listing the working processes utilizing CreateToolhelp32Snapshot and Process32Next.
0x3B6C27610D1 inject Decrypt and inject   shellcode, from encrypted type on disk, into reminiscence.
0xDA83E71 exec Execute a given command line.
0xE9478DC reside Get the PID of the method wherein the orchestrator is working.
0x6D6E70D40 cacls Modify the entry controls of a given object utilizing SetEntriesInAclA, SetNamedSecurityInfoA and BuildExplicitAccessWithName.

X4 offers fundamental functionalities to manage the machine remotely, however it lacks extra superior spying capabilities.


The LookBack backdoor has beforehand been described by Proofpoint; we’re due to this fact offering a fast abstract and our evaluation of the customized community protocol.


In all samples we noticed, the LookBack loader is a reputable model of libcurl.dll with the curl_share_init (ordinal #52) export modified to load the SodomNormal communications module. This corroborates the remark by Proofpoint researchers. This module is embedded within the library’s useful resource part and encrypted with an algorithm much like RC4. The encryption/decryption perform, proven in Determine 20, at all times makes use of the identical key.

Determine 20. Decompiled view of the perform used to encrypt and decrypt the embedded module

The SodomNormal element tries to learn configuration info from a sodom.ini file. This configuration file is encrypted utilizing the just-described perform and begins with the magic bytes 0xAF1324BC. If this file is unavailable or invalid, a hardcoded default configuration is used.

A novel sufferer ID is then generated from the sufferer’s CPUID, username, and IP deal with. That is despatched to the server together with the pc’s title and the configuration knowledge. The communications module then downloads the principle backdoor module, named SodomMain, from the C&C server. Sadly, we couldn’t receive this module.

Communication protocol

LookBack can talk over HTTP or by way of its “regular protocol”. In both case, the information being transferred is identical.

LookBack’s regular protocol makes use of uncooked TCP sockets and a customized message format described in Desk 4. This message consists of eight header fields, adopted by a physique of variable size. The message physique is encrypted with the perform beforehand described for the SodomNormal useful resource within the loader (Determine 20). The encrypted knowledge is then compressed with the deflate algorithm by way of the compress perform of the statically linked zlib.

Desk 4. LookBack message format

Subject Offset (bytes) Observe
Magic bytes 0x00 The fixed 0x48AB2EC2. Messages that don’t begin with this magic worth are discarded.
Compressed physique measurement 0x08
Uncompressed physique measurement 0x0C
Checksum 0x10 CRC32 of the message physique.
Message kind 0x14 Integer worth indicating the message’s content material and the related motion to be carried out.
We have now discovered code for over 50 message sorts. There appears to be little to no overlap between the values utilized by the consumer and the server. Desk 5 presents the categories we’ve got   analyzed in additional depth.
Message Physique 0x20 The message physique might be empty. On this case, the checksum and size fields are set to 0x00.

Desk 5. LookBack message sorts

Message kind Utilized by Description
2 Shopper Register with C&C server. The physique comprises configuration and details about the sufferer host.
3 Server Acknowledgment for message kind 2.
8 Shopper Request to obtain the principle backdoor element (SodomMain).
9 Server Reply to message kind 8. The message physique comprises the SodomMain file.
36 and 38 Shopper Switch file to server in message physique.
35 and 37 Server Response to message 36 or 38.
41 Shopper Request file from server.
42 Server Switch file to consumer in message physique (response to message 41)

The HTTP protocol makes use of the message format detailed within the earlier paragraph, however it provides just a few further steps to disguise its site visitors as reputable HTTP. It makes use of a pair of hardcoded templates, one for consumer requests and one other for server responses. The fields required for HTTP, corresponding to content material size, deal with, and port quantity, are stuffed in with the proper values; ineffective knowledge is used for the others.

For consumer requests, the messages are encoded with a modified hexadecimal algorithm that makes use of the encoding alphabet a-p as an alternative of the traditional 0-9a-f. This offers some obfuscation and ensures that messages won’t comprise binary knowledge or be clearly hex encoded, each of which may look suspicious in an utility/x-www-form-urlencoded message. The request’s physique consists of this encoded worth prefixed with the hardcoded string id=1&op=report&standing=. Shopper request and server response templates are proven in Determine 21 and Determine 22 respectively, with the template fields in angle brackets.

Determine 21. Template used for HTTP consumer requests

On the server facet, the information described within the earlier part is shipped straight as binary knowledge within the physique with a header purporting it’s a GIF picture.

Determine 22. Template used for HTTP server responses

TA410 – JollyFrog

This third staff makes use of off-the-shelf malware from the recognized malware households QuasarRAT and Korplug (aka PlugX). JollyFrog largely aligns with what was described by Fortinet as APT10.


Korplug, also referred to as PlugX, is a backdoor that has been used for years by many alternative cyberespionage teams. Regardless of being well-known, it’s nonetheless in use and we’ve got noticed TA410 utilizing it as just lately as in April 2021.

Within the case of TA410, Korplug arrives as a RARSFX archive, typically named m.exe, containing three recordsdata:

  • qrt.dll: A customized loader.
  • qrtfix.exe: A reputable signed utility from F-Safe, weak to DLL search-order hijacking.
  • qrt.dll.usb: The Korplug shellcode.

The loader allocates reminiscence utilizing VirtualAlloc and copies the content material of qrt.dll.usb there. Then it jumps proper into the shellcode that may decompress and cargo the Korplug payload.


QuasarRAT is a full-featured backdoor freely obtainable on GitHub. It’s utilized by quite a few menace actors who carry out cyberespionage or cybercrime.

TA410 makes use of a customized downloader and a customized loader written in .NET, that are handy for figuring out their situations of QuasarRAT amongst all of the noise created by different attackers.

Named sll.exe, this downloader is digitally signed with the certificates seen in Determine 23. The certificates is probably going stolen and belongs to 北京和赢讯时科技有限公司 (translated: Beijing Heyingxunshi Expertise Co., Ltd.) with thumbprint 850821D88A4475F0310F10FBA806353A4113D252. Though the certificates has now been revoked, it was nonetheless legitimate when this pattern was signed on August 10th, 2020.

Determine 23. Digital signature of the QuasarRAT downloader

This downloader merely downloads the loader and encrypted QuasarRAT payload from the hardcoded C&C server http://ffca.caibi379[.]com, at /rwjh/new/. This server was beforehand linked to FlowCloud (FlowingFrog). The loader is known as PresentationCache.exe and is protected with DNGuard, a business .NET packer. It is usually signed with the identical certificates because the downloader. It decrypts and hundreds the ultimate QuasarRAT payload, which makes use of cahe.microsofts[.]org as its C&C server.


TA410 is a cyberespionage umbrella concentrating on high-profile entities corresponding to governments and universities worldwide. ESET is revealing its newest findings about this group, together with outcomes from ongoing  analysis, throughout Botconf 2022.

Preliminary entry to targets is obtained by exploiting weak internet-facing purposes corresponding to Microsoft Trade, or by sending spearphishing emails with malicious attachments corresponding to RTF paperwork created by way of the Royal Street builder. Regardless that the JollyFrog staff makes use of generic instruments, FlowingFrog and LookingFrog have entry to complicated implants corresponding to FlowCloud and LookBack. YARA and Snort guidelines for these implants can be found in ESET’s GitHub repository.

For any inquiries about our analysis printed on WeLiveSecurity, please contact us at [email protected]

ESET Analysis now additionally affords non-public APT intelligence reviews and knowledge feeds. For any inquiries about this service, go to the ESET Threat Intelligence web page.



SHA-1 Filename Detection Description
C96558312FBF5847351B0B6F724D7B3A31CCAF03 N/A Win32/Agent.UWR FlowCloud v5.0.3 preliminary loader.
1403241C415A8D686B1148FA4229A2EB833D8D08 setlangloc.dll Win32/Agent.UNL FlowCloud DLL hijacking malicious library.
38D0E92AFF991CFC9C68D7BAAD6CB85916139AF5 hidmouse.sys Win32/Agent.UKR TA410 32-bit Rootkit/Keylogger driver.
AF978ED8AD37CE1437A6B42D96BF518D5C4CFD19 hidmouse.sys Win64/Agent.UKR TA410 64-bit Rootkit/Keylogger driver.
B70F3A3A9B5B8506EE95791469CA496E01AD7DAF winver32.dll Win32/Agent.ULH FlowCloud v4.1.3 hcClientLoaderZero_x86 backdoor.
014421BDB1EA105A6DF0C27FC114819FF3637704 hhh.exe Win32/Agent.ABYK FlowCloud v4.1.3 preliminary loader.
EA298866E5A61FEEA4D062987F23B10A78C8A4CA N/A Win32/Agent.ULH FlowCloud v4.1.3 backdoor.
021B9E2E8AA30B29569254C0378A9F43E4F32EEC winver64.dll Win64/Agent.KM FlowCloud v4.1.3 hcClientLoaderZero_x64 backdoor.
2A2F08FAD6B0A86DC94885224687D954E739CC21 N/A Win32/ParanoidFish.A Pafish sandbox detection instrument.
3658B7CCA13C8C8AD03E9B6AEFE4B9CBE48E3C81 hidmouse.sys Win64/Agent.UKR TA410 Rootkit/Keylogger driver.
517488F6BD0E7FC9EDE82F37226A75212B277E21 hidmouse.sys Win64/Agent.UKR TA410 Rootkit/Keylogger driver.
C05B4AD7A3322917E17710842FB88A090198D51F N/A Win32/Agent.TWI LookBack trojanized libcurl loader.
DB2DF1BDF8145CB8ABA3A2026A3CC3EF4F1762BE phx.dll Win32/Agent.TWI LookBack trojanized libcurl loader.
EDE2AB811311FC011B1E89C5A0B7A60C123B7398 hidmouse.sys Win64/Agent.UKR TA410 Rootkit/Keylogger driver.
7AA35BA7030AFCD271436DE8173D7B2F317A1BFC libcurl.dll Win32/Agent.TWI LookBack trojanized libcurl loader.
A5C02ABE698300F3DE0B7CC7F0856652753831DA libcurl.dll Win32/Agent.TWI LookBack trojanized libcurl loader.
613C4AFAE8F5F80F22DCD1827E3230FCA361ADA5 libcurl.dll Win32/Agent.UKD LookBack trojanized libcurl loader.
859CD6DFDADAB3D6427C6C1C29581CB2094D648F meterpreter.exe Win32/Rozena.CP Metasploit Meterpreter backdoor.
DBEA7F0C0D2BF8BC365A2D1572CA1538FE8FB9A3 responsor.dat Win32/Agent.ULL FlowCloud fcClientDLL ultimate stage backdoor.
ADD5B4FD9AEA6A38B5A8941286BC9AA4FE23BD20 絆邧坋蔡趕口昴.doc Win32/Exploit.Agent.TY Malicious Royal Street doc.
7BA42061568FF6D9CA5FE5360DCE74C25EA48ADA N/A Win32/Agent.ACKQ Packed Tendyron downloader.
D81215890703C48B8EA07A1F50FEC1A6CA9DF88B N/A Win32/TrojanDownloader.Agent.FLI Unpacked Tendyron downloader.
F359D3C074135BBCA9A4C98A6B6544690EDAE93D OnKeyToken_KEB.dll Win32/Injector.ELGA Tendyron malicious DLL.
621B31D5778EC2FB72D38FB61CED110A6844D094 N/A Win64/Rozena.AO X4 community shellcode.
BC11DC8D86A457A07CFE46B5F2EF6598B83C8A1F m.exe Win32/Injector.EMVA Korplug dropper.
C369E1466F66744AA0E658588E7CF2C051EE842F qrt.dll Win32/Injector.EMVA Korplug loader.
B868764C46BADC152667E9128375BA4F8D936559 qrt.dll.usb N/A Korplug encrypted payload.
BDECA89B4F39E6702CE6CBBC9E6D69F6BBAB01C8 N/A N/A Korplug decrypted payload.
5379FBB0E02694C524463FDF7F267A7361ECDD68 sll.exe MSIL/TrojanDownloader.Agent.GPS QuasarRAT downloader.
6CC6170977327541F8185288BB9B1B81F56D3FD0 PresentationCache.exe MSIL/Agent.TZG QuasarRAT loader.
D95185A4A3F8512D92F69D2ED7B8743638C54BE8 N/A MSIL/Spy.Agent.AES QuasarRAT backdoor.
BE7F0E41CD514561AED43B07AA9F5F0842BF876C HTra.exe Win32/HackTool.Hucline.AB HUC Packet Transmitter (HTran).
7F663F50E9D6376715AEB3AB66DEDE038258EF6C HTran13.exe Win32/HackTool.Hucline.S HUC Packet Transmitter (HTran).
BEDA1224B3BB9F98F95FF7757D2687F4D9F4B53A occasion.exe Win32/Agent.UJN Easy cmd.exe-based backdoor compiled with MingW.
2B61E7C63A0A33AAC4CF7FE0CEB462CF6DACC080 htran.exe Win32/HackTool.Hucline.AB HUC Packet Transmitter (HTran).
EF3C796652141B8A68DCCF488159E96903479C29 htran_f-secury.exe Win32/HackTool.Hucline.AB HUC Packet Transmitter (HTran).
6B547C244A3086B5B6EA2B3A0D9594BBE54AE06B inbt.zip Python/HackTool.Agent.J EXE masquerading as ZIP. This can be a Python community scanner (compiled with PyInstaller).
4CDCE3AF614C2A5E60E71F1205812AB129C0955B msd017.exe Python/Exploit.MS17-010.B This can be a Python scanner (compiled with PyInstaller) for the vulnerability MS17-010 (EternalBlue).


Serial quantity 0F8B600FF1882E
Thumbprint 02ED6A578C575C8D9C72398E790354B095BB07BC
Topic CN Hangzhou Leishite Laser Expertise Co., Ltd.
Topic O Hangzhou Leishite Laser Expertise Co., Ltd.
Topic L Hangzhou
Topic S Zhejiang
Topic C CN
Legitimate from 2012-03-29 09:07:04 UTC
Legitimate to 2014-04-02 06:24:19 UTC
Serial quantity 4ED8730F4E1B8558CD1CB0107B5F776B
Thumbprint 850821D88A4475F0310F10FBA806353A4113D252
Topic CN 北京和 赢讯时 科技有限公司 (translation: Beijing   Heyingxunshi Expertise Co., Ltd.)
Topic O 北京和 赢讯时 科技有限公司 (translation: Beijing Heyingxunshi Expertise Co., Ltd.)
Topic OU 研 发 部 ( R&D Division)
Topic S 北京市 (Beijing)
Topic C CN
Legitimate from 2019-11-13 00:00:00 UTC
Legitimate to 2020-11-12 23:59:59 UTC


Area IP First seen Particulars
43.254.216[.]104 2020-06 Supply server
45.124.115[.]103 2020-08 Supply server
161.82.181[.]4 2020-12 Supply server
43.254.219[.]153 2020-07 X4 C&C server
154.223.141[.]36 2020-06 HTran C&C server
103.139.2[.]93 2020-10 Tendyron C&C server
cahe.microsofts[.]com QuasarRAT C&C server
ffca.caibi379[.]com QuasarRAT downloader C&C server
smtp.nsfwgo[.]com Korplug C&C server
45.124.115[.]103 2020-06 LookBack C&C server
185.225.19[.]17 2021-01 LookBack C&C server
94.158.245[.]249 2020-03 LookBack C&C server
5.252.179[.]227 2021-03 LookBack C&C server
222.186.151[.]141 2019-11 FlowCloud C&C server
47.111.22[.]65 2020-09 FlowCross C&C server
114.55.109[.]199 2020-05 FlowCloud C&C server
dlaxpcmghd[.]com 185.225.17[.]39 2020-09 LookBack C&C server
wwww.dlmum[.]com N/A FlowCloud C&C server

MITRE ATT&CK methods

This desk was constructed utilizing version 9 of the MITRE ATT&CK framework.

Tactic ID Title Description
Useful resource Improvement T1587.001 Develop Capabilities: Malware TA410 develops LookBack and FlowCloud.
T1588.003 Get hold of Capabilities: Code Signing Certificates TA410 makes use of stolen code-signing certificates.
T1588.005 Get hold of Capabilities: Exploits TA410 had exploits for ProxyLogon and ProxyShell.
Preliminary Entry T1190 Exploit Public-Dealing with Utility TA410 has exploited net server vulnerabilities for preliminary entry.
T1566.001 Phishing: Spearphishing Attachment TA410 makes use of malicious RTF and DOCX attachments to compromise victims.
Execution T1106 Native API FlowCloud makes intensive use of the Home windows API to execute instructions and launch processes.
T1129 Shared Modules TA410’s backdoors can load DLLs and execute their payloads.
T1203 Exploitation for Shopper Execution TA410 makes use of Royal Street RTF paperwork to compromise victims.
T1559.001 Inter-Course of Communication: Part Object Mannequin FlowCloud makes use of COM interfaces to schedule duties and carry out WMI queries.
T1047 Home windows Administration Instrumentation TA410 makes use of WMI for lateral motion and data gathering.
Persistence T1053.005 Scheduled Process/Job: Scheduled Process FlowCloud creates a scheduled activity for persistence.
T1505.003 Server Software program Part: Internet Shell TA410 vegetation webshells on weak net servers.
T1543.003 Create or Modify System Course of: Home windows Service FlowCloud might be configured to create a service for persistence.
Protection Evasion T1027 Obfuscated Recordsdata or Info FlowCloud recordsdata are distributed and saved in encrypted type.
T1036.004 Masquerading: Masquerade Process or Service The motive force element of FlowCloud masquerades as a mouse driver service.
T1036.005 Masquerading: Match Reliable Title or Location Recordsdata named after reputable utilities are written into the %ProgramFilespercentMSBuildMicrosoftExpressionBlendmsole subdirectory.
T1014 Rootkit FlowCloud makes use of a rootkit to cover its community site visitors and processes from system utilities.
T1055.001 Course of Injection: Dynamic-link Library Injection FlowCloud makes use of each common and reflective DLL injection. It additionally manually hundreds some DLLs, bypassing calls to LoadLibrary.
T1055 Course of Injection TA410’s backdoors carry out course of injection to masquerade as innocent processes.
T1055.003 Course of Injection: Thread Execution Hijacking Certainly one of FlowCloud’s DLLs replaces directions within the loading course of to make it execute code written in its reminiscence.
T1055.012 Course of Injection: Course of Hollowing FlowCloud makes use of module stomping to cover the loading of its fundamental backdoor.
T1140 Deobfuscate/Decode Recordsdata or Info A number of TA410 backdoors talk with their C&C by means of encrypted and obfuscated channels.
T1574.002 Hijack Execution Move: DLL Facet-Loading FlowCloud makes use of DLL Facet-Loading to launch its second-stage dropper.
T1497 Virtualization/Sandbox Evasion Some variations of FlowCloud use the Pafish utility to detect virtualization, sandboxes, and debuggers.
T1134.002 Entry Token Manipulation: Create Course of with Token FlowCloud can create processes utilizing tokens acquired from reputable processes.
T1070.004 Indicator Removing on Host: File Deletion FlowCloud deletes its rootkit’s executable after launching it.
T1070.006 Indicator Removing on Host: Timestomp FlowCloud backdates some recordsdata and providers to 2013.
Discovery T1010 Utility Window Discovery When logging mouse occasions, FlowCloud gathers details about the appliance working within the foreground.
T1057 Course of Discovery A number of TA410 backdoors can record working processes.
T1518 Software program Discovery FlowCloud makes use of the IShellAppManager COM object to record put in software program.
T1083 File and Listing Discovery FlowCloud can search by means of related file methods and procure listing listings.
T1120 Peripheral Machine Discovery FlowCloud can record related digital camera units.
T1016 System Community Configuration Discovery FlowCloud can uncover and use regionally configured proxies.
T1012 Question Registry FlowCloud parts use registry keys to sign one another.
T1115 Clipboard Information FlowCloud registers a listener to steal clipboard knowledge when it’s modified.
Assortment T1056 Enter Seize FlowCloud logs mouse clicks.
T1056.001 Enter Seize: Keylogging FlowCloud information keystrokes.
T1113 Display screen Seize FlowCloud takes screenshots at common intervals.
T1125 Video Seize FlowCloud makes use of OpenCV to take photos utilizing related digital camera units.
T1123 Audio Seize FlowCloud has audio seize performance.
T1119 Automated Assortment FlowCloud robotically collects knowledge based mostly on timers and occasions.
T1074.001 Information Staged: Native Information Staging FlowCloud shops collected knowledge in native SQLite databases previous to exfiltration.
T1005 Information from Native System FlowCloud can exfiltrate recordsdata from native file methods.
T1025 Information from Detachable Media FlowCloud can exfiltrate recordsdata from detachable drives.
T1560.002 Archive Collected Information: Archive by way of Library FlowCloud and LookBack use a statically linked zlib library to compress knowledge.
T1560.003 Archive Collected Information: Archive by way of Customized Methodology FlowCloud compresses some collected knowledge by eradicating duplicates and related display screen captures.
Command And Management T1071.001 Utility Layer Protocol: Internet Protocols LookBack and FlowCloud can ship and obtain knowledge over HTTP.
T1095 Non-Utility Layer Protocol LookBack can talk over uncooked TCP sockets.
T1132.001 Information Encoding: Commonplace Encoding FlowCloud makes use of Protobuf to encode C&C instructions and configuration.
T1132.002 Information Encoding: Non-Commonplace Encoding LookBack encodes binary knowledge utilizing a customized hex-encoding methodology.
T1573.001 Encrypted Channel: Symmetric Cryptography FlowCloud can use XOR, TEA, RC4 and a modified AES algorithm to encrypt site visitors and recordsdata.
Exfiltration T1030 Information Switch Measurement Limits FlowCloud makes use of native caches to stage knowledge and exfiltrates their content material when it reaches a measurement laid out in its configuration.
Affect T1529 System Shutdown/Reboot FlowCloud can pressure a system crash or shutdown.

Posted in SecurityTags:
Write a comment