This blog was authored by Warren Mercer and Paul Rascagneres.

The malicious document in question is written in Korean with the following title:

5170101-17년_북한_신년사_분석.hwp (translation: 5170101-17 __ North Korea _ New Year _ analysis .hwp)

This document was alleged to be written by the Korean Ministry of Unification and included their logo as a footer on the document.

An interesting twist also came within the analysed malicious document as it attempts to download a file from an official Korean government website: (Korean Government Legal Service). The file downloaded is a binary masquerading as a jpeg file that is later executed as part of the infection. It's likely that the website was compromised by the attackers to try and legitimise the HTTP GET attempts for the final payload, this traffic would potentially not have looked unfamiliar for any system administrators.

The attackers' infrastructure appeared to be up for a few days at a time with no observed infrastructure re-use occurring. Unfortunately, the compromised sites were all either cleaned or removed by the attackers and Talos were unable to obtain the final payload. This level of operational security is common for sophisticated attackers.

Due to these elements it's likely that this loader has been designed by a well-funded group in order to target public sector entities in South Korea. Many of these techniques fit the profile of campaigns previously associated with attacks by certain government groups.

Infection Vector: Hangul Word Processor
The infection vector identified by Talos is a HWP file. This is a fairly unusual choice as this software is rarely used outside of Korea, but it is known to be widely used within Korea, including use by the South Korean government. As a regional file format, many security devices are not equipped to process HWP files. This can allow an attacker a vector with a much lower risk of detection by any security scanning devices.

Here is a screenshot of the opened document:

The title of the document is "Analysis of "Northern New Year" in 2017". The logo at the bottom of the document in the logo of the Ministry Of Unification. This ministry is working towards the reunification of North & South Korea. The document describes information linked to the North Korean celebration of New Year.

At the end of the document are 2 links to additional documents. The malicious document mentions that users should double click in order to access to these documents, Document1 is identified as "Comparison of Major Tasks in '16 & '17" and the Document2 linked is identified as "Comparison between '16 & '17"

The links point to 2 OLE objects embedded in the document (BIN0003.OLE and BIN004.OLE):

Once decompressed (zlib), we identified two PE32 files embedded within the 2 OLE files. If the targeted user double clicks on one of the links, a PE32 file is dropped and executed.

The 2 dropped binaries will be found and executed in this location during our analysis:

  • C:\Users\ADMINI~1\AppData\Local\Temp\Hwp (2).exe
  • C:\Users\ADMINI~1\AppData\Local\Temp\Hwp (3).exe We can identify a JavaScript object in the document. This one does not contain malicious content, it's an object included by default.

Here is the execution of the HWP file in Cisco AMP Thread Grid:

Dropped files
The compilation path of the binaries was not removed which allows us to determine the working space and environment used for this attack.

e:\Happy\Work\Source\version 12\T+M\Result\DocPrint.pdb

The two dropped malware files have a different hash but their purpose is the same:

  • Open a HWP document (to respond to the double click in the previous document)
  • Download a payload from a compromised host/C2.
    The opened document is embedded in the PE (in a resource named 'DOC'):

Like the previous document, this one speaks of the relation between North Korea and South Korea, and is seemingly written by a native Korean speaker due to the specific language used.

The second stage of the binary executes wscript.exe and injects shellcode into the process. The shellcode is embedded in a resource called 'BIN'. The purpose of this shellcode is to unpack a second PE32 in the legitimate wscript.exe process and execute it. The injection is perform by the classic:
VirtualAllocEx(), WriteProcessMemory() and CreateRemoteThread() APIs.

The unpacked binary is used to collect information on the infected system, and to attempt to communicate with the C2 in order to download the final payload. The information collected was:

  • The computer name
  • The username
  • The execution path of the sample
  • The BIOS model by analysing the HKLM\System\CurrentControlSet\Services\mssmbios\Data\SMBiosData registry key. This information allows the attackers to identify Virtual Machine (on VirtualBox the model is "innotek GmbH VirtualBox")
  • An ID randomly generated to identify the system
    This information could be used as a reconnaissance phase to determine if there was a suitable platform to deliver the final payload and to avoid sending the final payload to sandbox systems.

The analysed sample performed network connections to these 2 URLs in this order:

  • (where the collected information is sent)
    The beginning of the jpg document (02BC6B26) is the ID previously generated. We think that the jpg file is automatically generated by the index.php file if the collected data is relevant. The content of the jpg file is saved in a file called 'officepatch.exe'. Finally, this new file is executed and the unpacked executable terminates itself.

The website is the web site of the Korean Government Legal Service. Talos can only assume that this website was compromised in order to deliver the final stage malware, the jpg file. All the infrastructure was down during our analysis, which meant we were unable to analyse the payload directly.

The collected binaries are compiled between 22:43:05 UTC and 4:55:18 UTC (the 3 files at 22:00:00 are the binaries dropped by the HWP document and the other files are the unpacked payload) - Time Stamp artifacts can be easily faked and can be deployed as a false flag mechanism to make the researcher believe the compiled code came from a certain Time Zone - this should not be trusted as an indicator of where the attack or attacker originated from.

Command & Control infrastructure
During our investigation we were able to identify additional Command and Control infrastructure used by this actor. The four C2s were based in the following countries:

  • 3 C2 in South Korea
  • 1 C2 in the Netherlands Here is a global map of the identified infrastructure:

Colour Key:

  • Red: the '_put.jpg' binary (final payload)
  • Orange: C2 infrastructure used by the attackers
  • Yellow: the unpacked samples that perform the connection to download the final malware (the green bubbles share 90% of similar codes)
  • Green: the dropped executable by the HWP document (the orange bubbles share 90% of similar codes)
  • Blue: the HWP document

This actor appears to have made intentional decisions to limit the attack surface by using Hangul. This allowed them to evade some security devices as this format is not frequently processed.

The infection process was a MalDoc with multiple droppers (identical in their execution) and then C2 communication to obtain the final payload. The use of decoy documents is very common and shows that the attacker wanted to use a social engineering / enticement aspect to encourage the users to open the file.

This campaign has clearly targeted at a specific group of users, this rings true with the use of such specific file formats. Steps were clearly taken to limit the ability of security products to detect the threat as well as adherence to a strict timeline to prevent the malicious files from being discovered. The attackers were careful to remove their malicious payloads and not re-use their infrastructure.

We believe this is a targeted attack aimed at South Korean users in the public sector conducted by a sophisticated threat actor with access to native Korean speakers. Attacks on these individuals may be an attempt to gain a foothold into assets which can be deemed extremely valuable.


5170101-17년_북한_신년사_분석.hwp: 281828d6f5bd377f91c6283c34896d0483b08ac2167d34e981fbea871893c919

Dropped files:


Unpack related samples:

21b098d721ea88bf237c08cdb5c619aa435046d9143bd4a2c4ec463dcf275cbe 761454dafba7e191587735c0dc5c6c8ab5b1fb87a0fa44bd046e8495a27850c7
3d442c4457cf921b7a335c0d7276bea9472976dc31af94ea0e604e466596b4e8 930fce7272ede29833abbfb5df4e32eee9f15443542434d7a8363f7a7b2d1f00
4b20883386665bd205ac50f34f7b6293747fd720d602e2bb3c270837a21291b4 f080f019073654acbe6b7ab735d3fd21f8942352895890d7e8b27fa488887d08

Network: (where 02BC6B26 is randomly generated) (where 02BC6B26 is randomly generated) (where 02BC6B26 is randomly generated) (where 02BC6B26 is randomly generated)

Additional ways our customers can detect and block this threat are listed below.

Advanced Malware Protection (AMP) is ideally suited to prevent the execution of the malware used by these threat actors.

CWS orWSA web scanning prevents access to malicious websites and detects malware used in these attacks.

Email Security can block malicious emails sent by threat actors as part of their campaign.

The Network Security protection ofIPS andNGFW have up-to-date signatures to detect malicious network activity by threat actors.

AMP Threat Grid helps identify malicious binaries and build protection into all Cisco Security products.

Umbrella prevents DNS resolution of the domains associated with malicious activity.