Executive Summary

Netryx RAT has begun to gain public visibility as a new remote-access threat developed in Rust that combines traditional RAT capabilities with information theft, HVNC, persistence, remote execution, and an experimental integration with AI language models.

According to the publication record analyzed by iQBlack, the first observable release corresponds to Netryx RAT v0.2.2, published on November 4, which already advertised the extraction and decryption of browser passwords, cookies, and payment cards, support for Chrome App-Bound Encryption, keylogging, shell access, clipper functionality, and hidden desktop capabilities.

Four days later, v0.3.5 was being distributed with an architecture based on an operator panel, agent, and relay, Linux relay support through Docker, and new capabilities for organizing, searching, and managing compromised clients.

On November 9, the first independent public technical reference also appeared. Joe Security analyzed a sample identified as DeviceSync.exe and classified it as NETRYX RAT, confirming a Rust implementation, backdoor functionality, persistence mechanisms, and code associated with an LLM workflow. The analysis classified the sample as a backdoor and assigned it a score of 94/100.

That same day, source code attributed to Netryx v0.3.8 was obtained, distributed across several parts and accompanied by agent documentation and a changelog, substantially expanding visibility into the project’s internal architecture.

From Stealer to Remote Access Platform

The project’s chronology shows rapid expansion. Netryx v0.3.1 introduced a payload builder, persistence through different Windows mechanisms, execution of Python and Lua scripts, and improvements to HVNC and file transfers. The developer also acknowledged multiple functional issues and published fixes as users reported them.

With v0.3.3, released on November 6 within the sequence observed on Telegram, the project substantially expanded its information-stealing component. Its developers claimed to have extended the stealer to 53 target classes, including FTP clients, development tools, remote-access applications, email services, gaming platforms, messaging applications, VPNs, password managers, and cryptocurrency wallets.

The same version also introduced payload generation in .bat, .ps1, .vbs, and .hta formats, power controls, agent management, and additional panel functionality.

The source code subsequently released reinforces this characterization. The agent README describes a windowless Windows application with dedicated modules for HVNC, keylogging, screenshots, file management and downloads, credential theft, webcam access, persistence, process control, and reverse proxy functionality.

Its command surface additionally includes an interactive shell, process management, injection, HVNC, webcam capture, and operations involving files and persistence.

Everything indicates that Netryx is attempting to combine information collection, persistence, remote control, victim management, and interactive execution within a single platform.

Rapid, Unstable, and Now Reactive Development

The pace of evolution also provides insight into the threat’s maturity. Between November 4 and 8, versions v0.2.2, v0.3.0 “Initial public release (Skidrat rebrand)”, v0.3.1, v0.3.2, v0.3.3, and v0.3.5 were observed in succession, accompanied by frequent corrections involving the relay, builder, HVNC, panel, and agent management.

The project’s own messages acknowledge components that did not work correctly, results that were not being stored, connection errors, and functionality that was initially announced but subsequently required correction.

At the same time, the project promotes a development model open to its community: a bot exists for reporting bugs and submitting suggestions, installation tutorials are provided, and successive source-code packages have been distributed. The new release attributed to v0.3.8 adds another dimension to this process.

Its UPDATE_LOG.md documents versions v0.3.4 through v0.3.8 published on November 9 and describes the addition or restoration of modules related to privilege elevation, persistence, registry editing, anti-analysis, and techniques intended to interfere with EDR/AV solutions.

Particularly relevant is v0.3.7, which explicitly describes part of its changes as “Joe Sandbox hardening”. The developer claims to have applied information obtained from the analysis of DeviceSync.exe to modify the installation name, network transport, string obfuscation, behavior against sandboxes and debuggers, and variability between builds.

LLM Integration Already Exists in the Code

The feature that may distinguish Netryx most clearly at this early stage is the incorporation of an assistant based on language models.

Joe Security identified within DeviceSync.exe a workflow that allows an operator to provide a prompt, use an API key, select a model, and construct a request intended for an LLM service. The analysis also recovered a default model and the structures required to send the request and return the response content through the RAT’s result channel.

This provides independent evidence that the LLM layer does not exist solely as a promotional reference within the project’s ecosystem.

Joe Security did not observe successful LLM queries and found no evidence that generated responses are automatically forwarded into the command-execution mechanism. It also identified a possible weakness in HTTP response processing that could affect the current implementation.

For now, Netryx therefore should not be described as a RAT that allows an AI system to autonomously control compromised systems.

The available evidence supports a more precise conclusion: an LLM integration is compiled into the malware and accessible to the operator, but its complete operational use remains under development or, at minimum, has not yet been publicly demonstrated.

Analytical Closing

Netryx RAT remains at an early stage and shows clear signs of incomplete development. In only a few days, the project has publicly progressed from capabilities primarily associated with information theft toward an architecture incorporating persistence, HVNC, script execution, remote administration, a substantially expanded stealer, evasion mechanisms, and an experimental LLM-based layer.

Netryx is not only expanding its functional surface. Its developers also appear to be adapting the project as flaws and characteristics are identified by both users and malware-analysis tools.

Netryx still presents more questions than answers regarding adoption, distribution, victims, and operational effectiveness. Its evolution, however, already provides a relevant signal: what began around information-stealing capabilities is rapidly attempting to become a broader remote-access platform, while its developers add new functionality, experiment with artificial intelligence, and react to defensive intelligence produced about the malware itself.