Introduction Streams of malicious emails Talos inspects every day usually consist of active spamming campaigns for various ransomware families, phishing campaigns and the common malware family suspects such as banking Trojans and bots.. It is however often more interesting to analyze campaigns smaller in volume as they might contain more interesting malware. A few weeks ago I became interested in just such a campaign with a smaller number of circulating email messages. The email, first of them submitted from Middle East, purports to be coming from a Turkish trading company, which might further indicate the geographic area where the attacks were active. Analyzing malware is often like solving a puzzle, you have to do it piece by piece to reach the final image. In this case I spent more time analyzing the campaign than I initially planned. The campaign has many stages of the infection chain and all needed to be unraveled before the final payload level was reached. Furthermore, each of the stages used different development platform and was obfuscated in a different way. But let us start from the beginning.

Stage 1 - email
The email message contains two attachments. The first one is a Word document in the Office Open XML file format while the second is a ZIP file, containing an executable file PurchaseOrders.exe. This is a relatively unusual strategy for email campaigns as it is much more common for malicious emails to contain a single attachment rather than two or more. It seems that the attackers wanted to be double sure that the recipient will open at least one of the attachments.

Email campaign

Stage 2a - Word Document - CVE-2013-3906
The Word document attachment, “Letter of introduction.doc” contains an exploit for CVE-2013-3906 tiff image file parsing vulnerability. The document contains multiple TabStrip (classid: {1EFB6596-857C-11D1-B16A-00C0F0283628}) ActiveX controls also used in CVE-2012-1856.

Embedded ActiveX controls used for heap spray
Embedded TabStrip ActiveX controls are used for heap spraying and the embedded TIFF file named image1.jpeg triggers the CVE-2013-3906 vulnerability. There are 40 embedded ActiveX controls and each is mapped in 2MB allocated memory space. In this case, exploitation takes time but the exploit eventually crashes the vulnerable versions of Word and starts the shellcode. The shellcode is immediately visible in the hex dump of the ActiveX OLE2 file and sprayed in the memory of the exploited Word process.

Heap sprayed shellcode from ActiveXn.bin files

Shellcode - hook evasion The shellcode itself is relatively simple and, give or take, 450 bytes long, excluding the URL used for downloading the payload. As is often the case, the APIs are found by parsing the Process Environment Block (PEB) and traversing the linked list of loaded modules as well as their respective exported functions.

Notably, before calling required APIs, the shellcode checks for presence of inline hooks, often installed by endpoint security products and jumps over the installed hook code in order to avoid being noticed in their behavior detection windows.

Evading security hooks  
If the user was infected by the attached Word document, the shellcode would download and execute an executable from a legitimate, compromised server. The C2 server for the final payload is extracted from a configuration blob stored encrypted within the downloaded payload body.

Stage 2b - PurchaseOrders.exe The executable downloaded by the shellcode is identical in its functionality to the executable attached to the email so we are eventually coming to PurchaseOrder.exe which will eventually get executed whether the user opens the attached document or if they immediately go for launching the PurchaseOrder.exe. The executable has a PDF document icon and the user can be forgiven for not recognizing it as an executable, considering the fact that Windows by default hides filename extensions of the known file types.

Icon file used by PurchaseOrder.exe. The executable itself is just over 1.4MB in size, which is rather large for attachments used in email campaigns. The file itself is a self-extractable CAB archive which contains three randomly named files.

Stage 3 - AutoIt Script The first file is instantly recognisable and it is a legitimate, Autoit script interpreter. The second file is a Unicode file encoded as UTF-16 and is over 110MB in size which is at first almost enough to discourage from analysis. The actual script code starts deep within the file, which provides the attacker with the ability to obfuscate the script code in a way that is not immediately visible by researchers.

Talos has already written about a similar delivery method in the past and it seems that this campaign uses a similar generator of obfuscated Autoit scripts. Thankfully, It was relatively simple to remove all the junk characters and reduce the size of the code to analyze to a much more manageable 41KBs.

Autoit stage deobfuscated
The Autoit script itself creates a directory in the user's profile folder and sets its attributes to system and hidden. It then creates a copy of RegSvcs.exe .NET services installation tool or copies the existing RegSvcs.exe to a filename splwow64.exe to set up the next stage. Regscvcs.exe is used for injecting and launching a remote thread within its process space. The thread uses RC4 to decrypt the third file dropped by the original self-extractable CAB archive and reads it into the process space of regsvcs.exe. This leads us to the next stage, using an executable developed in C/C++. This stage will only exist in its executable format in memory, while it will be an RC4 encrypted data blob on the disk.

Stage 4 - Zyklon injector The stage injected into RegSvcs.exe is another unobfuscated injector of the final payload. The executable decompresses the payload from the resource section of the PE file, finds and launches Windows Explorer executable that is found in different folders depending on the Windows platform (32 or 64 bit) and launches a remote thread that loads and runs a .NET executable, which is the final payload of the campaign, in this case a sample of the Zyklon HTTP bot.

Loading managed code into an unmanaged space is not entirely simple process. Attempting to cheat the infection chain to launch the Zyklon bot from the command line was apparently anticipated by the campaign author who modified the Zyklon class Main function to display a text message for anybody trying to launch it this way.

The original Zyklon code for the version does not seem to contain this mechanism that ensures that the payload is run by a specific loader that does not call the Zyklon Class Main function but a different entry point.

You are not supposed to run it this way
The payload is obfuscated using Crypto Obfuscator and an additional code generator. The code which uses xor operations to set a value of a variable used in a switch statement to direct the program flow is relatively easy to follow once the Crypto Obfuscator code transformations are removed, which can be done using a very useful .NET deobfuscation utility de4dot. In fact, the Zyklon Builder, found on VirusTotal, uses the same dnlib library, used also by de4dot and dnspy analysis tools, to add the configuration file to the malicious .NET assembly base Zyklon bot embedded in its resource section.

Once the obfuscator was removed it did not take too long to realize that for the purpose of the analysis it was possible to manually modify the Zyklon class Main function to call the EntryPoint function which contains the bot code and debug the Zyklon using the dnspy debugger.

Establishing communication with a C2 server
The bot connects to one of the three possible C2 servers, starting from the first one specified in its configuration. The server sends a certificate and the communication is first encrypted with RSA and then with a 256 bit long AES with the initialization vector and the key generated by the server, sent back to the client after the client POSTs a request ending with the query gate.php?getkey=y.

Looking at the DNS requests for one of the C2 servers that remained active throughout the campaign it is possible to see the time when the campaign was active.

C2 DNS domain activity
The initial configuration for the bot is embedded within the resource section of the file, together with the list of user agent strings used by the bot when contacting the C2 server. The malicious .NET assembly also contains an encrypted blob that becomes its persistence module injector. Once decrypted and loaded in memory its function is to make sure that the bot is respawned from a remote thread if the main executable is terminated as a process.

The client then sends a request containing the information about the infected system and receives a configuration string from C2, which sets the internal bot parameters. Several threads are also launched in order to download and execute required additional plugins.

The main command loop sleeps for 60 seconds and sends a request for a command to the C2 server. The main purpose of the bot seems to be conducting DDoS attacks but there are other more or less standard commands available such as downloading and executing additional payloads from a user-specified URL or logging the user keystrokes and sending them back to the C2 server.

Curiously, Zyklon may also attempt to enumerate the usual automatic startup locations in the Windows registry to find potential competitive files and submit them to VirusTotal for scanning. So called cloud malware inspection is used to terminate processes based on the VirusTotal verdict. The bot also executes rudimentary heuristic checks for some of the known competitive bot names and filename extensions and tries to remove them if found on the system. Competition is never welcome by the bad guys.

Zyklon website
The website advertising Zyklon is hosted on a .onion domain which is also accessible from the clear net through a web to Tor proxy. The owners are advertising two different versions for sale, one that can connect to Tor based C2 servers and the standard one without that capability.

Perhaps the most interesting page of the Zyklon website are its Terms of Service, which the authors seem to believe may free them from potential prosecution. The user, aka the attacker, allegedly has the sole legal responsibility for damage caused by it, at least according to Zyklon creators :


1. use the Zyklon H.T.T.P Remote Administration Software for any illegal purpose, or in violation of any laws, including, without limitation, laws governing intellectual property, data protection and privacy, and import or export control;

2. remove, circumvent, disable, damage or otherwise interfere with security-related features of the Zyklon H.T.T.P Remote Administration Software, features that prevent or restrict use or copying of any content accessible through the Zyklon H.T.T.P Remote Administration Software, or features that enforce limitations on use of the Zyklon H.T.T.P Remote Administration Software;

3. intentionally interfere with or damage operation of the Zyklon H.T.T.P Remote Administration Software or any user's enjoyment of them, by any means, including uploading or otherwise disseminating viruses, worms, or other malicious code;

4. post, store, send, transmit, or disseminate any information or material which infringes any patents, trademarks, trade secrets, copyrights, or any other proprietary or intellectual property rights;or

5. Install and/or use Zyklon H.T.T.P Remote Administration Software on any computer which you do not have explicit permission to do so on;

6. distribute Zyklon H.T.T.P files over the Internet with the intent of infecting/harming machines of other people;

Downloaded credential harvesting modules (email, browser, ftp) Zyklon creators also advertise a number of useful plugins for harvesting user credentials and stealing confidential information such as details of wallets of various crypto currencies like Bitcoin, LiteCoin and DodgeCoin. For a potential customer, the list of features must be quite impressive. However, not everything is as ideal as it seems at first.

In the analyzed campaign, the Zyklon main executable downloaded only three plugins, as instructed by the C2 server, all of them with a purpose of stealing user credentials from password caches of the most popular web browsers as well as email and ftp clients.

CI=False|KT=1|UAC=False|S5=False|ER=False|UPNP=False|RP=True|RW=False| AK=False|BK_CYCLE=|BK_RUN_ONCE=False|SOCKS_PORT=3128|SOCKS_AUTH=False| SOCKS_USERNAME=Nothing|SOCKS_PASSWORD=Nothing|KLI=1|KLM=500|EKL=True| WC=False|BA=MyBtc|LA=MyLtc|KLF=False|BR=True|FTR=True|EMR=True|SFR=False| GR=False|AU=False|UF=N/A|

Configuration command sent to Zyklon from C2 server
The plugin download URL follows the format of plugin/index.php?plugin=<pluginname> with possible plugins being


Available Zyklon plugins
Downloaded plugins are injected into a previously launched and hollowed legitimate process name "%windir%\Microsoft.NET\Framework\v4.0.30319\RegAsm.exe" and are in fact just freeware command line tools written in C/C++ available from the website

It is likely that Zyklon author realized it would take quite a long time to fully develop all the features within the main Zyklon bot and decided to include available free password dumping utilities just to make its RAT more competitive in what is quite a cutthroat underground market for remote administration tools.

Conclusion Zyklon is quite a well known botnet kit and it has been fairly active this year. In this smaller, possibly more targeted campaign we analyzed, it has shown that its users are employing a number of different technologies and obfuscation techniques to be more successful - from exploiting a vulnerability in Microsoft Word over Autoit scripts and .NET executables, all the way to freeware utilities used as plugins for harvesting credentials from browser cache, email and ftp clients.

Zyklon campaign execution flow on an endpoint
Overall, this was a well executed campaign which used compromised hosts as C2 servers. Luckily, there are several weaknesses which can be exploited for detecting its footprint either by inspecting IOCs or tracking the network communications patterns and behavior on endpoints.


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

CWS or WSA 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.

Network Security appliances such as NGFW, NGIPS, and Meraki MX with Advanced Security can detect malicious activity associated with this threat.

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.

Stealthwatch detects network scanning activity, network propagation, and connections to CnC infrastructures, correlating this activity to alert administrators.

Iocs Document exploits

PE Exes downloaded by the exploit docs

ZIP Attachments

PE exe files within ZIPs

Browser plugin

FTP plugin

Email plugin

Zyklon Builder v1.0.0

Download URLs

C2[.]php[.]php - Control Panel