Skip to content
NewCybercrime Report 2026 · Eight Chapters · A Post Is Not An Incident → Explore The ChaptersCybercrime Report 2026 → Explore
Security NewsNews Explainer

Operation ShadowCloud: How GitDoor and GraphDoor Use Cloud Services

Operation ShadowCloud used MeshCentral Agent to deploy GitDoor and GraphDoor. Follow the observed chain, cloud channels, and checks for defenders.

By
Brandefense Research Team
Reviewed by
Brandefense Analyst Team
Published
18 November 2026
Reading time
7 min read
Two abstract cloud paths branch from one remote access node.

Operation ShadowCloud is an active intrusion campaign that Kaspersky GReAT identified in early 2026. Its researchers found victims in Asia and Africa, primarily in government networks. The attackers used a legitimate remote management agent to place two distinct backdoors, GitDoor and GraphDoor, on compromised Windows systems.

The important boundary is the one between an approved management tool and the code it later delivered. Kaspersky could not establish how the agent first arrived. Its evidence suggests that an earlier compromise may have been reused, so the agent’s presence alone does not identify the entry point.

How Does Operation ShadowCloud Move From Remote Access to a Backdoor?

At one investigated organization, Kaspersky found two instances of the MeshCentral Agent. The agent provides legitimate remote desktop, command line, and file management functions. In this case, an adversary used the established remote session to place a trusted executable, a malicious loader library, and an encrypted payload on an endpoint.

The trusted program then loaded the attacker’s library. Kaspersky observed several legitimate binaries used for this DLL side loading technique. The loader decoded the encrypted payload and ran the backdoor in the current process’s memory, which complicates file based inspection.

Observed ShadowCloud chain
  1. Existing remote sessionMeshCentral Agent provides access; initial entry remains unknown
  2. Trusted executableA malicious library is loaded alongside it
  3. LoaderThe encrypted payload is decoded and run in memory
  4. BackdoorGitDoor or GraphDoor contacts a cloud service
  5. Plug-insAdditional capabilities are fetched when required

This sequence matters because each step can look ordinary in isolation. A remote support agent may be authorized, and the executable it launches may be signed. For Cyber Threat Intelligence teams, the suspicious relationship is an unexpected agent session followed by unusual library loading and a new cloud communication pattern.

What Separates GitDoor From GraphDoor?

Both implants gather host information, accept commands, and download modules on demand. They also use cloud platforms for command and control, making simple network blocking less reliable. Their service choices and command channels differ, so investigations should distinguish them rather than treat the names as aliases.

GitDoor uses GitHub repositories to receive instructions and return results. Kaspersky also observed it sending collected system information through Ably. The implant records context such as the host name, user name, operating system, running process, and network addresses before exchanging encrypted data.

GraphDoor uses Microsoft Graph to access a SharePoint document library. Kaspersky found it authenticating as an application, placing encrypted host data in the library, and retrieving plug-ins and commands through the same service. Its plug-ins send status back to the parent process through an internal communication channel.

Two cloud command paths
FunctionGitDoorGraphDoor
Primary cloud serviceGitHub repositoriesMicrosoft Graph and SharePoint
Additional channelAbly messagingAbly based proxy modules
Shared behaviorHost profiling and modular commandsHost profiling and modular commands

Neither GitHub nor SharePoint traffic is malicious by itself. The investigative question is whether a newly launched process has a legitimate business reason to make these connections. Security teams can pair endpoint ancestry with cloud audit data rather than relying on service names alone.

Why Do the Plug-ins Change the Risk?

The implants can fetch extra capabilities after the initial backdoor starts. Kaspersky describes modules for keylogging, process listing, screen capture, and network proxying. This allows an operator to adapt activity to the target without putting every function in the first executable.

The proxy components are especially relevant to government network defenders. Kaspersky found a modified Erfrp client and Ably based channels that can relay traffic. That design can extend access beyond the first host, although the public research does not establish that every observed victim experienced the same follow-on activity.

A keylogger or screen capture module changes the response scope as well. If either ran, teams should assess exposed credentials and information visible in affected sessions. That is an investigative priority derived from the modules’ functions, not a claim that Kaspersky documented a specific stolen account.

The campaign also illustrates why cyber threat intelligence needs to connect malware behavior with local telemetry. A cloud destination alone provides weak context. Process lineage, the reason for remote agent access, and the sequence of downloaded capabilities provide a stronger basis for triage.

What Can Defenders Check Along the Chain?

Start with approved remote management inventory. Identify where MeshCentral Agent is allowed, who can start sessions, and which endpoints recently gained an instance. Review installation and session records for the affected period, then compare them with change tickets and administrator activity.

Next, inspect child processes and library loads around those sessions. Kaspersky’s chain pairs a legitimate executable with an unexpected library and encrypted payload. An endpoint investigation should preserve that relationship, including parent process, signer, file origin, and time, before files are removed.

Then correlate cloud activity from the same host and process. GitDoor points investigators toward GitHub and Ably usage; GraphDoor points toward Microsoft Graph and SharePoint. Search for unusual application authentication, new repository or document library operations, and traffic from processes with no approved reason to use those services.

Finally, review evidence of later capability use. Look for unexpected keyboard hooks, screen capture, process enumeration, or proxy behavior on the host. These are behavior classes, not unique ShadowCloud indicators, so analysts should require a coherent sequence before escalating a case.

ShadowCloud investigation order
  • Validate remote accessCompare agent sessions with the approved inventory and change records.
  • Preserve process lineageRecord the executable, loaded library, payload origin, and timestamps.
  • Correlate cloud eventsMatch endpoint activity with GitHub, Ably, Graph, and SharePoint logs.
  • Scope follow-on accessReview credential exposure and other hosts reached through remote sessions.

If the sequence is present, isolate the affected endpoint under the organization’s response procedure and preserve evidence. Revoke unauthorized remote sessions and review associated application credentials. Scope other hosts touched by the same account or management path before declaring containment.

What Does the Research Establish About Attribution?

Kaspersky says it lacks enough information to definitively attribute Operation ShadowCloud to a specific group. It observed GitDoor staged in directories also used by implants associated with FamousSparrow and CopperTurtle. On several systems, the deployments were within one day of each other.

That overlap is a clue, not proof of a shared operator. Kaspersky suggests that a Chinese-speaking group may be developing or using the backdoors, while leaving the actor unnamed. The threat actor label should remain provisional in intelligence records and incident reports.

The distinction matters operationally. A hunt based on the demonstrated chain can begin now, even if the actor identity changes. Attribution can inform hypotheses, but it should not narrow the initial search to a known group’s older infrastructure or techniques.

What Remains Uncertain for an Investigation?

The initial access method remains unknown in Kaspersky’s public account. The research documents two MeshCentral Agent instances at one organization and a wider campaign across Asia and Africa. It does not give a public count of affected organizations or show that every victim received both backdoors.

Similarly, a modular capability is not proof that it ran on every host. Investigators should separate a module available to the operator from evidence of its execution locally. This prevents a possible exposure from becoming a reported fact about stolen data or lateral movement.

The campaign is best treated as an observed access and command chain with several possible follow-on actions. Use the source’s sequence to guide collection, then let endpoint and cloud logs determine the scope of a specific incident. A later attribution assessment can follow the evidence rather than precede it.

Frequently Asked Questions

How did Operation ShadowCloud first enter the networks?

Kaspersky did not establish the initial entry point. Its evidence suggests an earlier infection may have been reused before MeshCentral Agent deployed the payloads.

Are GitDoor and GraphDoor the same malware?

No. Kaspersky describes distinct backdoors with shared modular capabilities. GitDoor uses GitHub for commands, while GraphDoor uses Microsoft Graph and SharePoint.

Does SharePoint or GitHub traffic confirm a ShadowCloud infection?

No. Investigators need the related remote agent activity, process lineage, and unusual cloud operations to assess whether a host matches the reported chain.

Who operates Operation ShadowCloud?

Kaspersky says definitive attribution is not possible from its evidence. It describes staging overlap with FamousSparrow and CopperTurtle activity as a possible link.

Sources

  1. Kaspersky Securelist: Initial Infection and GitDoor
  2. Kaspersky Securelist: GraphDoor, Plug-ins, and Attribution
Share this articleLinkedInXEmail
Brandefense logo

Stay Ahead of Digital Threats.

Take control of your digital security with an exclusive demo of our AI-driven threat intelligence platform.

Open The Full Demo Form →