Malware Turns Microsoft 365 Calendar Into Covert Attack Vector

Security researchers at Group-IB have uncovered a Windows malware strain that turns Microsoft 365 calendars into covert channels for receiving commands and stealing files from targeted organizations. The malware component, dubbed HOLLOWGRAPH, abuses the Microsoft Graph API and a compromised Microsoft 365 account to conceal command-and-control activity within legitimate cloud communications.

The malware uses the compromised account's calendar as what Group-IB called a "two-way dead drop." Attackers create calendar events containing encrypted instructions while HOLLOWGRAPH uploads stolen files as attachments to separate events.

Those events are scheduled for May 13, 2050, keeping them far outside the mailbox owner's normal calendar view. The approach allows the malware to communicate through Microsoft infrastructure without connecting directly to an attacker-controlled server for its primary command channel.

HOLLOWGRAPH supports only two commands: "get" for retrieving instructions and "send" for exfiltrating files. The limited command set suggests the malware operates as one part of a larger collection of tools rather than as a standalone backdoor.

Group-IB linked the malware with high confidence to the Cavern backdoor framework based on similarities in command syntax and architecture. The firm found at least 12 infected systems, though only three were actively communicating with the attacker during the observation period.

The earliest communication occurred June 3 and the latest was recorded July 9. A compromised mailbox tied to an Israeli organization and malware samples uploaded from Israel indicate that the campaign has focused primarily on Israeli targets.

The small number of infections suggests the operation is "highly targeted rather than opportunistic," Group-IB said.

HOLLOWGRAPH also maintains a secondary communication method using Domain Name System tunneling. The malware sends specially constructed IPv6 address queries to an attacker-controlled domain to obtain updated Microsoft Entra ID credentials, including the tenant ID, client ID, client secret, and mailbox address needed to access Microsoft Graph.

Commands and stolen data sent through the calendar are protected with RSA and AES encryption. Separate key pairs handle incoming instructions and outgoing files, preventing exposure of one key from immediately compromising both communication directions.

Group-IB said it could not confidently attribute the campaign to a known threat actor. Researchers found some technical overlap with Lyceum, an Iranian-linked subgroup of OilRig, but assessed that connection with low confidence. The targeting and technical design nevertheless point to a "capable and well-resourced adversary," the firm said.

The campaign extends a broader trend in which attackers hide malicious activity behind trusted services and familiar business tools. Recent malware operations have abused signed applications, remote management software, and cloud infrastructure to blend into legitimate enterprise activity.

Group-IB recommends that security teams monitor Microsoft Graph and mailbox audit logs for unusual calendar operations, particularly events dated far into the future. Defenders should also look for events with encrypted text attachments, application-generated subject changes, and the logAzure.txt configuration file.

Orgs should audit OAuth applications, rotate exposed credentials and watch for unusually frequent IPv6 DNS queries or long, high-entropy subdomains that could indicate tunneling activity. For more information, check out Group-IB's full technical analysis here.

About the Author

Chris Paoli (@ChrisPaoli5) is the associate editor for Converge360.

Featured

  • Robot typing on fluorescent keyboard

    The Wrong Battle: Why Your Institution's AI Policy Is Probably Solving the Wrong Problem

    The conversation on most campuses has become consumed with detection: How do we catch students using AI when they shouldn't? The impulse to protect academic integrity is legitimate, but the detection-first approach has a fatal flaw.

  • Abstract futuristic digital network with glowing padlock icons

    Microsoft Intros New Agentic AI Security Multi-Model Defense System

    A new multi-model agentic AI security system built by Microsoft's Autonomous Code Security team helped researchers find 16 new vulnerabilities across the Windows networking and authentication stack, the company anounced in a recent security blog post.

  • Abstract digital data stream with binary code and colorful light trails

    Microsoft Releases Open Source AI Safety Tools for Agent Development

    Microsoft released RAMPART and Clarity as open-source projects intended to help developers test AI agents earlier in the software lifecycle and turn red-team findings into repeatable engineering checks.

  • abstract smartphone translucent screen displaying AI interface

    Apple Introduces Redesigned Siri AI

    At its recent Worldwide Developers Conference, Apple introduced Siri AI, a redesigned version of its voice assistant that Apple describes in its own announcement as "a profoundly more capable and personal assistant." The update is intended to make Siri more conversational, more context-aware, and more useful across iPhone, iPad, Mac, Apple Watch, and Vision Pro.