Last seven days
- First activity
- Aug 31, 2026
- Last activity
- Aug 31, 2026
- Feed role
- C2 / Distribution
- Host form
- 1 IP / 0 hostnames
SmartLoader is a Windows malware loader used in large-scale fake GitHub repository campaigns, including the FakeGit and AgentBaiting operations.
Profile source: Mallory opens in a new tabSmartLoader
SmartLoader is a Windows malware loader used in large-scale fake GitHub repository campaigns, including the FakeGit and AgentBaiting operations. It is commonly delivered through trojanized ZIP archives embedded in cloned or impersonated developer projects, especially AI tools, developer utilities, game cheats, and Model Context Protocol or AI skill lures. Typical delivery packages contain a batch launcher, a renamed LuaJIT interpreter or runtime components, and a heavily obfuscated Lua payload disguised as a text or log file. Execution of the launcher starts the Lua-based stager, allowing the malware to blend into seemingly legitimate project files and evade simplistic file-based inspection.
SmartLoader is characterized by its LuaJIT-based staging architecture, extensive obfuscation, and dynamic command-and-control discovery. Multiple analyses associate its first-stage scripts with Prometheus-style Lua obfuscation, while later stages have shown indicators consistent with MoonSec. The malware uses anti-debugging and anti-tamper logic, suppresses visible console windows, fingerprints the host, performs internet and geolocation checks, and captures screenshots. It can exfiltrate host metadata and screenshots to its operators and uses the Polygon blockchain as a dead-drop resolver for command-and-control information, enabling operators to rotate infrastructure without rebuilding the malware.
Persistence is commonly established through Windows scheduled tasks, including redundant recovery paths that can relaunch a cached local stage or re-download encrypted stages from GitHub. SmartLoader also retrieves additional encrypted payloads from attacker-controlled GitHub repositories and contains functionality consistent with in-memory loading of follow-on payloads. In observed campaigns, SmartLoader has been used to deliver information stealers including StealC, Lumma Stealer, Vidar Stealer, and a previously reported NodeJS-based malware-as-a-service infostealer. Follow-on payloads have targeted browser credentials, cookies, cryptocurrency wallet data, email-related data, tokens, and other sensitive information.
The malware has been heavily associated with abuse of trust in open-source ecosystems and developer workflows. Campaigns using SmartLoader have impersonated legitimate repositories, copied documentation and project metadata, and leveraged public AI capability catalogs and AI assistants to increase discovery of malicious projects. Victims have been concentrated in North America, Asia, and Southern Europe, with notable targeting of financial services, banking, and technology organizations, although the lures also target individual developers and consumers.
C2 tracking
Derp observations, rolling seven-day window
Samples
0231fab970a4773956a60a93a613a58a0d5ca1fab2737a97eb591d635cfc44f3 3fd1b43c2c06c569cebd4865e4c29aac2088df774aaa00fbd96e34503495eab6 8f1712b4b01e63e948057b003ed9f03170b7e07ef00bd42c727c1566655a167c d0472c2cbdc7fbdd5a9448bda1a2df6127d4bf704b2dec080a536e92e8f9d893 ea27fa689b84ea7b60036e5fa33b8bd274f4d859f7beed73ac633fc052c8ed16 Reported operators
Together, the four files make up SmartLoader, a malware loader Netskope has previously seen distributed through GitHub repositories.
There has been an extensive malware campaign, dubbed FakeGit, that utilizes thousands of counterfeit GitHub repositories to distribute SmartLoader malware... Upon activation, the attack launches a LuaJIT-based loader that launches an obfuscated Lua script to install SmartLoader.
A newly documented campaign called AgentBaiting uses fraudulent AI Skills and Model Context Protocol, or MCP, servers to distribute SmartLoader malware through trusted-looking GitHub projects and public capability catalogs.
MITRE ATT&CK
Reporting
A malvertising campaign dubbed FakeAgent abused searches for the Claude Desktop app to deliver the SectopRAT information-stealing trojan to Windows users at at least 29 organizations. Victims who clicked sponsored Bing results were sent to a spoofed public artifact hosted on Anthropic's claude.ai domain instead of the legitimate download page, then redirected to attacker-controlled infrastructure that served a fake ClaudeDesktop.exe installer. Huntress said the activity ran from July 21 to 22 and should be treated as a full remote-access and credential-theft compromise, not a nuisance adware incident. The malware chain used DLL sideloading through a tampered libcef.dll, signed-binary proxy execution, VMProtect packing, GPU-based anti-analysis, and DirectX shader-based payload decryption to evade detection. Researchers said SectopRAT established persistence with sslconf.exe, tempdir.dll, and appcfg.dat under AppData, and retrieved command data via blockchain infrastructure tied to contract 0xc1907d7be91f95903ad66d775c397302e7dd9228; Huntress also identified 2.24.131[.]246 as a live command-and-control address. The RAT is capable of stealing browser credentials, cookies, autofill and payment data, Chromium keys, FTP credentials, Discord and messaging data, files, and passwords, and the infrastructure was linked to earlier malicious activity dating to May 2025, including domains associated with campaigns tied to StealC infrastructure seized during Operation Endgame.
Cisco Talos reported that the Chaos ransomware group is using a new Rust-based remote access trojan, msaRAT, to conceal command-and-control traffic inside legitimate browser activity. Instead of making direct outbound connections, the malware launches a headless Google Chrome or Microsoft Edge instance with the Chrome DevTools Protocol enabled, injects JavaScript, and uses the browser to establish a WebRTC DataChannel for encrypted communications. Talos said the malware combines WebRTC DTLS with its own ChaCha-Poly1305 encryption using an ECDH-derived shared key, giving operators covert tunneling, browser-assisted remote code execution, and asynchronous multi-threaded control. The reported intrusion chain begins with email or voice phishing, followed by the use of remote management software for persistence and delivery of an MSI installer disguised as a Windows update. That installer, identified as update_ms.msi, is downloaded over plain HTTP on port 443, extracts lib.dll, and loads it directly into memory before starting the browser in remote debugging mode. For signaling, the browser communicates with a Cloudflare Workers endpoint, while Twilio TURN infrastructure relays WebRTC traffic, a design that obscures attacker-controlled systems and makes network-based detection and attribution more difficult.
Derp wouldn't exist without the work these projects do for the security community. We rely on their data sources to improve the quality and depth of what we publish. Thank you, we're genuinely grateful.