ESET researchers have identified new MuddyWater activity primarily targeting organizations in Israel, with one confirmed target in Egypt. MuddyWater, also referred to as Mango Sandstorm or TA450, is an Iran-aligned cyberespionage group known for its persistent targeting of government and critical infrastructure sectors, often leveraging custom malware and publicly available tools. In this campaign, the attackers deployed a set of previously undocumented, custom tools with the objective of improving defense evasion and persistence. Among these tools is a custom Fooder loader designed to execute MuddyViper, a C/C++ backdoor. Several versions of Fooder masquerade as the classic Snake game, and its internal logic includes a custom delay function inspired by the game’s mechanics, combined with frequent use of Sleep API calls. These features are intended to delay execution and hinder automated analysis. MuddyViper enables the attackers to collect system information, execute files and shell commands, transfer files, and exfiltrate Windows login credentials and browser data. The campaign also leverages credential stealers (CE‑Notes and LP‑Notes) and reverse tunneling tools (go‑socks5), long a favorite of MuddyWater operators.
Although this is our first public blogpost covering MuddyWater, ESET researchers have been tracking the group for several years and have documented its activities in multiple ESET APT Activity Reports. Unlike previous campaigns of MuddyWater, which were often noisy and easily detected, the one covered in this blogpost demonstrates a more focused, sophisticated, and refined approach.
Key points of this blogpost:
- MuddyWater developers adopted CNG, the next-generation Windows cryptographic API, which is unique for Iran-aligned groups and somewhat atypical across the broader threat landscape.
- The group also used more advanced techniques to deploy MuddyViper, a new backdoor, by using a loader (Fooder) that reflectively loads it into memory and executes it.
- We provide technical analyses of the tools used in this campaign, including MuddyViper, the Fooder loader, the CE-Notes browser-data stealer, the LP-Notes credential stealer, the Blub browser-data stealer, and go‑socks5 reverse tunnels.
- During this campaign, the operators deliberately avoided hands-on-keyboard interactive sessions, which is a historically noisy technique often characterized by mistyped commands.
MuddyWater group overview
MuddyWater is a cyberespionage group active since at least 2017, primarily targeting entities in the Middle East and North America. It is one of the most active Iran-aligned APT groups tracked by ESET researchers and has links to the Ministry of Intelligence and National Security of Iran.
The group was first introduced to the public as MuddyWater by Unit 42 in 2017, whose description of the group’s activity is consistent with ESET’s profiling – a focus on cyberespionage, the use of malicious documents as attachments designed to prompt users to enable macros and bypass security controls, and a primary targeting of entities located in the Middle East.
Notable past activities include Operation Quicksand (2020), a cyberespionage campaign targeting Israeli government entities and telecommunications organizations, which exemplifies the group’s evolution from basic phishing tactics to more advanced, multistage operations; and a campaign targeting political groups and organizations in Türkiye, demonstrating the group’s geopolitical focus, its ability to adapt social engineering tactics to local contexts, and reliance on modular malware and flexible C&C infrastructure.
Besides its frequent activity, MuddyWater operations are often noisy. The group is known for its persistent targeting of government, military, telecommunications, and critical infrastructure sectors, typically using custom malware and publicly available tools to gain access, maintain persistence, and exfiltrate sensitive data. In addition to targeting its archenemy, Israel, the group appears to be targeting countries that maintain, or seek to strengthen, diplomatic ties with Iran.
ESET has documented multiple campaigns attributed to MuddyWater that highlight the group’s evolving toolset and shifting operational focus. While the earlier operations relied on broad targeting and relatively unsophisticated techniques, more recent campaigns demonstrate signs of technical refinement and increased precision.
In March and April 2023, MuddyWater targeted an unidentified victim in Saudi Arabia by deploying a batch script that downloaded a PowerShell-based backdoor, which was used to download and execute arbitrary payloads and subsequently to remove the initial payload from disk.
The group conducted a campaign in January and February 2025 that was notable for its operational overlap with Lyceum (an OilRig subgroup), further detailed in this publication. This latest overlap suggests an evolution in MuddyWater’s modus operandi.
The group’s publicly documented custom tools include, for example, the Bugsleep, Blackout, Small Sieve, Mori, and POWERSTATS backdoors, as well as custom-compiled variants of open-source tools such as LaZagne or CrackMapExec. MuddyWater campaigns typically do not leverage or introduce new tools, malware, or techniques; instead, they are often noteworthy due to the targeting.
While MuddyWater initially concentrated strictly on cyberespionage, its cooperation with Lyceum led to targeting of the manufacturing sector through spearphishing. The attack generated considerable noise and achieved little in terms of operational objectives.
The campaign outlined in this publication shows what, for MuddyWater, seems to be an unprecedented advancement in toolset and technical execution.
Victimology
As previously mentioned, during this campaign, MuddyWater primarily targeted organizations in Israel, but also one in Egypt. Table 1 lists the victims by country and vertical. The campaign began on September 30th, 2024 and concluded on March 18th, 2025.
Table 1. Victims by country and vertical
| Country | Vertical |
| Egypt | Technology |
| Israel | Engineering #1 |
| Engineering #2 | |
| Engineering #3 | |
| Local Government #1 | |
| Local Government #2 | |
| Manufacturing | |
| Technology | |
| Transportation | |
| Utilities | |
| University #1 | |
| University #2 | |
| University #3 | |
| Unidentified #1 | |
| Unidentified #2 | |
| Unidentified #3 | |
| Unidentified #4 | |
| Unidentified #5 |
One interesting thing to note about the victim in the utilities vertical is that they were also compromised by Lyceum on February 11th, 2025.
Overlap and cooperation with Lyceum
In early 2025, ESET Research identified an operational overlap between MuddyWater and Lyceum, a subgroup of the Iran-aligned OilRig cyberespionage group, also known as HEXANE or Storm-0133. OilRig has been active since at least 2014 and is commonly believed to be based in Iran.
Tools associated with Lyceum include DanBot, Shark, Milan, Marlin, Solar, Mango, OilForceGTX, and various downloaders that use legitimate cloud services for C&C communication. Lyceum has targeted Israeli organizations, including government entities and healthcare organizations.
During a recent campaign, MuddyWater collaborated with OilRig in January and February 2025. They gained access through a spearphishing email with a link to an installer for Syncro RMM software. They then installed PDQ RMM tool and a custom Mimikatz loader disguised as certificate files. This allowed Lyceum to assume control of operations within a manufacturing organization in Israel.
The collaboration between MuddyWater and OilRig suggests MuddyWater may be acting as an initial access broker for other Iran-aligned groups.
The TTPs and tools used in this campaign align with capabilities attributed to MuddyWater, including the use of PowerShell and Go-based backdoors. MuddyWater’s targeting typically focuses on telecommunications, government, and energy sectors. They use spearphishing emails containing links to RMM software installers hosted on file-sharing platforms to gain initial access.
Code overlaps were identified between newly documented tools and those previously attributed to MuddyWater. These include credential stealers, backdoors, and customized reverse tunnels. The tools attempt to steal user credentials and use the CNG API for data encryption and decryption.
Newly documented tools used by MuddyWater include Fooder loader and MuddyViper backdoor. Fooder masquerades as the Snake game and uses a delay function to hide malicious behavior. MuddyViper allows attackers to collect system information, download/upload files, execute shell commands, and steal credentials.
Overall, the toolset used in this campaign includes Fooder loader, MuddyViper backdoor, CE-Notes, LP-Notes, Blub, and go‑socks5 reverse tunnels. Fooder loader is a 64-bit C/C++ tool that loads the MuddyViper backdoor into memory. It decrypts and loads the payload reflectively, allowing it to execute without writing to disk. The MuddyWater group has been known to use go-socks5 independently of Fooder, showing a consistent dependence on this tool for covert network communication and data exfiltration. Fooder, which often disguises itself as the Snake game, utilizes a custom delay function and Sleep API calls to slow down its malicious activities, as seen in Figure 3. Despite lacking built-in persistence capabilities, Fooder can enable persistence for the MuddyViper backdoor, a C and C++ backdoor that allows unauthorized access to compromised systems. MuddyViper sends detailed status messages and reports on security tools detected in the environment, resulting in significant network traffic. The backdoor supports various commands, including opening reverse shells, downloading/uploading/executing files, stealing user credentials, and setting up persistence. Passes the provided argument to the running reverse shell. Two C&C servers have been observed: processplanet[.]org and 35.175.224[.]64.
Both directions of communication AES-CBC encrypt the data, using the CNG API with the key (used across samples) 0608101047106453101617106423101013101012101083109710108585106969 and the IV 0.
In the backdoor → server direction of the communications:
– Each endpoint URI supported by the C&C server can be used by the backdoor for a specific type of request, such as requesting a command, uploading a file, or sending a custom status message.
– Additional data for the C&C server is included in the HTTP request body, which is unconventional for HTTP GET requests.
– The User-Agent string is A WinHTTP Example Program/1.0, a remnant of the example code for the WinHttpOpen API.
– The connection, send, receive, and response timeouts are set to 30 seconds.
– Default sleep time between consecutive connection attempts is 60 seconds. This value can be configured by command ID 700.
– Upon failure, connection attempts are retried up to 10 times.
– Prior to encryption, the data is always formatted as
In the server → backdoor direction of the communications:
– The HTTP status code determines the backdoor command ID.
– The backdoor command arguments are included in the HTTP response body.
If successful, the stolen credentials are then encrypted using AES-CBC with the CNG API, using the key ED15C8344B45DAED1E0578F8BC1A32411812C61F4CB45D89B107287DE0E09FFC and the IV 91A4E6F6D51DAEE773A8F00279792578.
Similar to CE-Notes, LP-Notes saves the encrypted credentials in a local file at C:\Users\Public\Downloads\lp-notes.txt. Another component is likely responsible for exfiltrating the data, such as an RMM tool or MuddyViper.
Blub Browser-Data Stealer
Blub is a C/C++ browser-data stealer that targets Google Chrome, Microsoft Edge, Mozilla Firefox, and Opera. It utilizes a statically linked SQLite library and is named after its executable, Blub.exe. The PDB path observed was C:\Users\jojo\source\repos\stealer\x64\Release\stealer.pdb.
Chromium-Based Browsers
For Chrome, Blub decrypts user credentials by obtaining and decrypting the encryption key from C:\Users\<username>\AppData\Local\Google\Chrome\User Data\Local State, using the CryptUnprotectData API. It then decrypts credentials from Chrome user profiles stored at C:\Users\<username>\AppData\Local\Google\Chrome\User Data\<profile_name>\Login Data.
A similar process is followed for obtaining and decrypting credentials from Microsoft Edge and Opera user profiles.
Firefox
Blub decrypts user credentials from Mozilla Firefox by parsing values from the logins.json file in each user’s profile directory.
The stolen data is stored in a local file named file.txt without encryption and is also logged onto the console with no encryption.
Blub checks for security solution processes before executing its payload, focusing on processes related to Avast and Norton.
go‑socks5 Reverse Tunnels
MuddyWater’s go‑socks5 reverse tunnels are tools compiled in Go, based on libraries like go‑socks5, yamux, and resocks. These tools are used to relay communication between compromised machines and C&C servers, enhancing stealth and communication security.
The go‑socks5 proxy acts as a communication relay between the compromised machine and the C&C server, using SSL/TLS authentication with a hardcoded connection key.
Conclusion
This campaign showcases MuddyWater’s evolving operational maturity, with the introduction of new components like Fooder loader and MuddyViper backdoor. The group demonstrates enhanced stealth, persistence, and credential theft capabilities, alongside game-inspired evasion techniques and reverse tunneling.
ESET will continue to monitor MuddyWater’s activities, focusing on technical advancements and strategic targeting across various sectors.
For inquiries about our research, contact us at threatintel@eset.com.ESET Research offers private APT intelligence reports and data feeds. Rewrite the sentence: “The cat chased the mouse around the house.”Rewritten sentence: “Around the house, the mouse was chased by the cat.” sentence: The cat chased the mouse around the house.
Rewritten sentence: The mouse was chased around the house by the cat. sentence in a different way:
Please rewrite the sentence in a different manner. sentence: Can you please pass me the salt?
Rewritten: Could you hand me the salt, please? Please rewrite the following sentence for me. Rewrite the sentence: “The cat chased the mouse around the house.”
Rewritten sentence: “Around the house, the mouse was chased by the cat.” sentence: Please make sure to lock the door before you leave. sentence: Could you please assist me with this task? sentence: The cat sat on the windowsill and watched the birds outside.
