"What makes SilkParasite interesting is the traces of AI-assisted development running through otherwise expert code, which is a different thing from AI-generated malware," Bitdefender Labs said in a technical report shared with The Hacker News.
SilkParasite's toolkit: seven RAT families, five newly documented
Security vendor Bitdefender Labs describes SilkParasite as an espionage intrusion set observed targeting government bodies in Central Asia. The cluster uses seven remote access tool (RAT) families — DriveSilkRAT, CookiETagRAT, NomadRAT, GoginRAT, NodeEdgeRAT, plus two previously known implants — and Bitdefender says five of those RATs have never been documented before.
The seven implants span four programming languages: .NET, C++, Go, and JavaScript. They are modular and plugin-oriented, allowing operators to expand capabilities without replacing core components. Bitdefender recorded roughly 65 instances infected with DriveSilkRAT, most of them located in the Asia region.
How the RATs operate: architecture and command channels
Bitdefender provided brief technical descriptions of each new family: DriveSilkRAT (.NET/C++) uses Google Drive as a command-and-control (C2) channel, polling a folder for tasking and returning results to the same location; it supports 12 plugins for process listing, enumeration, file management, and command execution. CookiETagRAT (C++) hides C2 in HTTP Cookie and ETag response headers. NomadRAT (C++) implements a main orchestrator and a transmitter library to handle C2 traffic and fetch plugins from the server by numeric identifiers only when required. GoginRAT (Go) shares architectural similarities with NomadRAT, uses a separate transmitter for C2, and routes plugin execution results through a shared callback. NodeEdgeRAT (JavaScript) packages command execution, file management, and transfer in a single script.
Nearly every tool in the set relies on DLL sideloading as the primary delivery vector. The operators bring a legitimately signed program, place a rogue DLL beside it under a name the application expects, and run the signed executable from an unusual location to trigger execution of the malicious library.
Audit-ready is a season. It shouldn't be.
Evidence in spreadsheets, controls drifting between audits, frameworks multiplying on flat headcount. Nubivance runs continuous compliance on Rapid7 Cyber GRC - SOC 2, HIPAA, ISO 27001, PCI, CMMC.
End the scrambleDelivery chain and regional targeting
Attack chains begin with password-protected RAR archives that contain malicious Microsoft Office documents. The password to open the archive is included in the email body — a common spear-phishing technique. Opening the document launches a macro that triggers a DLL sideloading sequence to drop the first-stage payload.
Bitdefender said the lures were regionally tailored: recovered documents were crafted to appear relevant to government entities in Uzbekistan, Turkmenistan, Kyrgyzstan, Tajikistan, and Kazakhstan, with several impersonating specific ministries. One recovered document, posted to a public malware-sharing platform, was addressed to a Georgian government entity. The campaign’s macros also check whether Kaspersky antivirus is installed and running before proceeding — a sign of deliberate attempts to bypass detection given the product’s prevalence in the region.
Signals of AI assistance and reuse of legacy Chinese tool components
Bitdefender distinguishes AI-assisted development from AI-generated malware. The clearest AI signal is an "indubitably AI-generated" phishing lure; other traces of AI assistance appear in code artifacts and repeated high-level designs implemented in different languages. For example, GoginRAT contains Go test functions and a hard-coded AES key literal set to "0123456789abcdef," while NodeEdgeRAT carries a configuration field for an encryption key set to the literal "change_this_key." Bitdefender also notes NomadRAT and GoginRAT share a similar architecture implemented twice in two languages — a pattern the vendor links to AI-assisted workflows.
Several elements point toward a China nexus with medium confidence. SilkParasite reuses or updates tooling tied to China-linked clusters: the backdoor BLOODALCHEMY — an updated version of Deed RAT, itself a successor to ShadowPad (an evolution of PlugX) — and an updated SpiceRAT variant attributed to a Chinese-speaking actor codenamed SneakyChef. Elastic Security Labs first documented BLOODALCHEMY in October 2023 as part of REF5961 attacks against government organizations in Southern and Southeast Asia; the backdoor supports host data collection, file overwrites, and uninstall commands and is launched via a DLL loader sideloaded by a legitimate binary.
What this means for governments, technologists, and regional defenders
- Governments in Central Asia: Ministry impersonation and regional tailoring mean public-sector endpoints and email gateways should treat inbound documents with heightened suspicion and validate origins of password-protected archives rather than relying on apparent source labels.
- Technologists and security teams: Bitdefender highlights that the most consistent detection surface is the pairing of a signed application loading a colocated library while running from an unusual location. The vendor warns volume-based signature detection will miss low-footprint plugin implants that use legitimate cloud services; it recommends behavioral baselines that flag unusual relationships between processes and network services.
- Policymakers and incident responders: SilkParasite is the third prominent actor reported striking Central Asia in recent years after UAC-0063 and FamousSparrow, and the reuse of components like BLOODALCHEMY and updated SpiceRAT emphasizes the persistent reuse and evolution of toolsets across campaigns.
SilkParasite combines traditional espionage tradecraft — spear-phishing, DLL sideloading, and modular implants — with signs of modern development processes that include AI-assisted workflows. That hybrid approach complicates straightforward attribution while underscoring Bitdefender’s central recommendation: defenders who build behavioral detection of anomalous process–service relationships will be best positioned to spot these low-footprint operations as they adapt and reconfigure their modular toolsets.
