Executive Overview
In the continuously evolving landscape of enterprise cyberthreats, state-sponsored actors and sophisticated cybercriminal syndicates are shifting away from noisy, easily detectable infrastructure in favor of “Living off the Cloud” (LotC) tactics. By leveraging trusted enterprise platforms, native APIs, and business-critical software, threat actors can blend seamlessly into the background hum of corporate network traffic.
A stark illustration of this paradigm shift has recently been uncovered by threat intelligence researchers at Group-IB. Security analysts have unmasked a sophisticated, highly targeted Windows malware strain dubbed HOLLOWGRAPH. This novel tool turns a cornerstone of modern corporate productivity—the Microsoft 365 calendar—into a covert, encrypted command-and-control (C2) channel and data exfiltration conduit.
By abusing the legitimate Microsoft Graph API and exploiting compromised Microsoft 365 (M365) accounts, HOLLOWGRAPH bypasses traditional perimeter defenses. Instead of communicating with external, attacker-controlled servers to pull instructions or dump pilfered documents, the malware communicates entirely through cloud-native infrastructure. It achieves this by stashing encrypted instructions inside calendar events scheduled decades into the future and uploading stolen corporate data as hidden email attachments.
Group-IB’s investigation highlights a worrying trend: enterprise environments are increasingly being weaponized against themselves. With early indicators tying the malware framework to known threat groups and a clear geographic concentration targeting Israeli organizations, HOLLOWGRAPH represents an evolutionary leap in stealthy, targeted cyberespionage. This in-depth report examines the mechanics of the malware, its operational infrastructure, the strategic threat actor landscape, and the comprehensive defensive mitigations required to combat cloud-based abuse.
Detailed Chronology and Technical Mechanics
The discovery of HOLLOWGRAPH stems from meticulous incident response and threat hunting operations conducted by Group-IB researchers, who traced the initial telemetry of the campaign back to early summer.

The Timeline of Discovery
- Early June (June 3, 2024): The earliest recorded communication traffic linked to the HOLLOWGRAPH campaign appears. Compromised systems begin silently interacting with M365 infrastructure, utilizing deeply buried calendar entries to poll for instructions.
- Early July (July 9, 2024): The latest active communication channel is logged during the observation window. Security telemetry flags anomalous API calls and anomalous DNS tunneling queries originating from targeted networks.
- Mid-July to August 2024: Group-IB correlates malware samples uploaded from Israel with compromised mailboxes tied to Israeli organizations, confirming a localized, high-value targeting pattern.
Anatomy of a Cloud-Based "Dead Drop"
The core innovation of HOLLOWGRAPH lies in its novel exploitation of Microsoft 365 calendars as what researchers have termed a "two-way dead drop."
Traditional malware relies on hardcoded IP addresses or Domain Name System (DNS) lookups to connect directly to command-and-control servers operated by hackers. Security appliances, Endpoint Detection and Response (EDR) tools, and next-generation firewalls are specifically tuned to flag these outbound connections to unknown or suspicious external endpoints. HOLLOWGRAPH circumvents these defenses entirely by utilizing the legitimate Microsoft Graph API. Because enterprise environments routinely allow internal applications to query Microsoft Graph for productivity and administrative tasks, the malware’s traffic looks entirely benign to local monitors.
The mechanics of the communication loop operate as follows:
- Command Retrieval ("Get"): The threat actors access the compromised M365 account and create calendar events containing encrypted instructions meant for the malware.
- The 2050 Time-Anchor: To keep these malicious instructions completely hidden from the mailbox owner and internal IT auditors, the events are deliberately scheduled for May 13, 2050. Because this date is decades away, the events never trigger pop-up reminders, nor do they appear in the user’s standard daily or weekly calendar views.
- Data Exfiltration ("Send"): When the HOLLOWGRAPH agent on an infected host needs to exfiltrate stolen files or reconnaissance data, it packages the payload, encrypts it, and uploads the data as an attachment to a separate calendar event within the compromised account.
Remarkably, HOLLOWGRAPH supports a drastically limited command set consisting of only two core functions: "get" for retrieving instructions and "send" for exfiltrating files. This minimalist architectural footprint strongly implies that HOLLOWGRAPH is not a standalone backdoor meant for broad initial access. Instead, it functions as a modular component within a larger, more complex collection of post-exploitation tools.
Secondary Communications: DNS Tunneling and Entra ID Credential Harvesting
While the Microsoft 365 calendar serves as the primary operational channel for instructions and data theft, HOLLOWGRAPH maintains a robust secondary communications method to ensure resilience: DNS tunneling.

If the primary calendar channel is disrupted or if the malware needs to refresh its authentication tokens, it resorts to specialized DNS queries. The malware crafts custom IPv6 address queries and sends them directly to an attacker-controlled domain. Embedded within these DNS queries are high-entropy subdomains and encoded requests designed to bypass standard network filtering.
Through this DNS tunneling vector, the malware retrieves updated Microsoft Entra ID (formerly Azure Active Directory) credentials. These configuration parameters include:
- The target Tenant ID
- The Client ID
- The Client Secret
- The specific mailbox address required to authenticate and interact with the Microsoft Graph API
Cryptographic Layering
To ensure that intercepted calendar entries or network logs do not expose the contents of the campaign, HOLLOWGRAPH employs a robust cryptographic architecture. Both commands sent down through the calendar and data exfiltrated back to the attackers are rigorously protected using a combination of RSA and AES encryption.
Crucially, the malware utilizes separate key pairs to handle incoming instructions versus outgoing files. This cryptographic separation ensures that even if an analyst or automated defensive system manages to compromise or expose one key pair, it will not immediately compromise both directions of the attacker’s communication stream.
Supporting Context & Metrics: The Threat Landscape
While the absolute numbers of infections observed in this campaign are relatively small, the metrics tell a compelling story about the nature of modern targeted cyberattacks.

Scope and Scale of the Campaign
- Total Infected Systems Identified: Group-IB confirmed at least 12 infected systems globally during their investigation.
- Active Communication Window: Only 3 of the 12 systems were actively communicating with the attacker-controlled cloud infrastructure during the strict observation timeframe.
- Geographic Concentration: The discovery of a compromised mailbox tied to an Israeli organization, alongside malware samples uploaded from IP spaces within Israel, points to a highly focused geographic campaign.
Group-IB’s analytical assessment stresses that the small number of infections is a feature, not a bug. Rather than casting a wide net via phishing emails or automated exploit kits (opportunistic attacks), the deployment of HOLLOWGRAPH is characterized as "highly targeted." The threat actors invested significant effort into compromising specific enterprise accounts and maintaining covert access over extended periods without triggering broad alerts.
The Rise of "Living off the Cloud" and Living off the Land
HOLLOWGRAPH sits squarely within a broader, deeply concerning evolution in cyberattack methodologies. For decades, security teams focused primarily on blocking malware binaries, known malicious signatures, and unauthorized external connections.
In response, advanced adversaries have adapted by embracing "Living off the Land" (LotL) and "Living off the Cloud" (LotC) techniques. By abusing legitimate enterprise tools—such as signed administrative scripts, remote management software, legitimate cloud infrastructure, and APIs like Microsoft Graph—attackers erase the functional boundary between malicious activity and normal business operations.
When a security tool inspects an outbound HTTPS connection to Microsoft’s cloud API, it registers as trusted Microsoft traffic. When an internal application accesses a calendar entry, it registers as standard productivity behavior. This fundamental challenge—the difficulty of distinguishing between administrative convenience and malicious abuse—is precisely what makes cloud-based attack vectors like HOLLOWGRAPH so potent.
Attribution and Threat Intelligence Assessments
Determining "whodunit" in cyberspace is notoriously difficult, particularly when threat actors deliberately leverage third-party cloud infrastructure to mask their operational footprints. Group-IB’s analysis of the HOLLOWGRAPH campaign provides nuanced insights into the potential orchestrators behind the malware.

The Cavern Backdoor Framework Connection
With high confidence, Group-IB researchers linked the HOLLOWGRAPH malware strain to the broader Cavern backdoor framework. This assessment was made based on striking similarities in command syntax, structural architecture, and operational methodologies between HOLLOWGRAPH and previously documented Cavern deployments.
The Lyceum/OilRig Connection
When attempting to drill down to a specific nation-state or financially motivated syndicate, Group-IB exercised a conservative approach. The firm stated that it could not confidently attribute the campaign to a known, named threat actor.
However, researchers did note notable technical and behavioral overlaps with Lyceum (also known as Desert Falcon), an Iranian-linked subgroup frequently associated with the broader OilRig (APT34) threat collective. While Group-IB assessed this specific connection with low confidence, the convergence of targeting vectors (specifically focusing on Middle Eastern entities, particularly in Israel) and the technical sophistication of the malware design point directly toward a capable and well-resourced state-backed adversary.
Iran-linked APT groups have a documented history of pioneering novel covert communication channels, utilizing custom PowerShell scripts, DNS tunneling, and abusing legitimate cloud infrastructure to sustain long-term espionage campaigns against geopolitical rivals.
Future Outlook and Defensive Recommendations
As cloud adoption scales universally across global enterprises, adversaries will inevitably continue to pivot their operations into Software-as-a-Service (SaaS) and Platform-as-a-Service (PaaS) environments. The HOLLOWGRAPH campaign serves as a critical warning shot for Chief Information Security Officers (CISOs), Security Operations Center (SOC) directors, and cloud architects: perimeter defense is no longer enough.

Protecting an organization against threats that live inside legitimate cloud services requires a fundamental reimagining of visibility, auditing, and access control.
Comprehensive Threat Mitigation Strategies
To safeguard Microsoft 365 environments and enterprise networks against sophisticated cloud-abuse frameworks like HOLLOWGRAPH, security teams should immediately implement the following hardening and monitoring protocols:
1. Advanced Microsoft Graph and Audit Log Monitoring
- Anomalous Calendar Auditing: Security teams must actively monitor Microsoft Graph API activity and mailbox audit logs for abnormal calendar operations. Specifically, SOC analysts should write detection rules that flag calendar events dated far into the future (such as the 2050 time-anchor identified in this campaign).
- Attachment Inspection: Monitor for calendar events that contain encrypted text attachments, unusual file types, or repeated programmatic modifications.
- Application-Generated Subject Changes: Track instances where calendar item subject lines or metadata are modified programmatically by enterprise applications rather than human users.
- Configuration File Detection: Watch endpoints and cloud repositories for the presence of the specific
logAzure.txtconfiguration file utilized by the malware framework.
2. Rigorous OAuth Application Governance
- Audit Connected Apps: Organizations must perform comprehensive, continuous audits of all OAuth and enterprise applications registered within Microsoft Entra ID. Unused, legacy, or overly permissive application permissions must be immediately revoked.
- Principle of Least Privilege: Ensure that applications utilizing the Microsoft Graph API are granted only the absolute minimum permissions required for their intended business function (e.g., restricting calendar read/write scopes strictly to necessary service accounts).
3. Credential Hygiene and Rotation
- Proactive Credential Resets: Regularly rotate exposed or long-lived API secrets, client secrets, and tenant credentials. Implement automated expiration policies for service principal secrets to limit the lifespan of compromised keys.
- Phishing-Resistant MFA: Enforce strict, phishing-resistant Multi-Factor Authentication (MFA) across all administrative and user accounts, paying special attention to accounts with access to mailboxes and Graph API configurations.
4. Network and DNS Telemetry Analysis
- Detecting DNS Tunneling: Security engineering teams should implement advanced network monitoring to detect anomalous DNS traffic patterns. Specifically, watch for unusually frequent IPv6 DNS queries, abnormally long subdomains, and high-entropy domain requests that could indicate secondary data exfiltration or credential retrieval via tunneling.
Conclusion
The emergence of HOLLOWGRAPH marks a sobering milestone in cloud-era cyberespionage. By weaponizing everyday collaboration tools like Microsoft 365 calendars, sophisticated adversaries have demonstrated that enterprise infrastructure can be seamlessly subverted into an invisible command network. Defending against these advanced, low-noise threats requires security leaders to abandon reliance on simple perimeter defenses and embrace deep, contextual cloud auditing, rigorous OAuth governance, and proactive threat hunting tailored to catch the silent abuses happening right in plain sight.
