Sending Logs, Alerts, and Telemetry Through a Data Diode

Find Out How
We utilize artificial intelligence for site translations, and while we strive for accuracy, they may not always be 100% precise. Your understanding is appreciated.

Microsoft DART Investigates a SharePoint Breach Involving Two Separate Threat Actors

A multi-stage intrusion built for stealth, and the lessons every SharePoint organization should take from it.
By Oana Predoiu
Share this Post

Quick summary: DART (Microsoft's Detection and Response Team) investigated and published a ransomware intrusion report, actively targeting internet-facing, on-premises SharePoint servers. During the investigation, a second threat actor was found operating in the same environments simultaneously, suggesting an overlap in attack campaigns.

What Happened

Threat actor identified as China-based Storm-2603 has been targeting SharePoint servers since (at least) mid-2025. The attack exploited two known vulnerabilities:

  • CVE-2025-49706: improper Microsoft Office SharePoint authentication allows attackers to perform spoofing over a network
  • CVE-2025-49704: code injection in Microsoft Office SharePoint, allowing attackers to execute code over a network

Post-intrusion, Storm-2603 searched for additional entry points through reconnaissance activity, including requests for sensitive configuration files. In this case, win.ini and web.config.

The attack was executed in three stages:

1. The group used Velociraptor with SYSTEM-level privileges to map the environment, deliberately blending malicious activity with trusted administrative behavior.

2. Remote access was then established through multiple channels: Cloudflare tunnelling, Zoho Assist, and SSH connections configured through Visual Studio Code.

3. Privilege escalation followed, with new local and domain administrator accounts created to maintain access.

As DART investigated, they uncovered a second, unrelated threat actor operating in parallel. This actor used two techniques not associated with Storm-2603.

  • DLL side loading, to place malicious code inside legitimate Windows system files, so it executes under the cover of a trusted process
  • Custom backdoors; purpose-built tools designed to maintain hidden access that standard security tools are unlikely to recognize

Both techniques are engineered for stealth, obscuring the presence of the second actor. At the time of writing, the second actor has not been identified.

One attacker getting in is bad enough.

Two, operating independently, in the same compromised environment, is a different problem entirely, suggesting that the moment a vulnerability is disclosed, multiple threat actors will race to exploit it. Storm-2603 may have been the first to exploit the two CVEs, but they weren't the last; and they almost certainly won't be the only group going forward.

The report doesn’t confirm the two actors were connected or coordinating. But intentionally or not, they were obscuring each other via two sets of malicious activity layered into the same environment.

Who Is Affected

Based on Microsoft’s report, the intrusion affected at least two organizations. Forensic evidence of lateral movement led DART to identify and engage a second victim, which was confirmed to have been compromised by the same ransomware activity as the first.

No further organizations have been confirmed as compromised yet.

However, any organization running on-premises SharePoint servers that has not applied Microsoft's security updates for CVE-2025-49706 and CVE-2025-49704 remains exposed to the same attack vector.

Microsoft Recommendations

In their report, DART recommended a series of remediation steps and security policies, addressing the full attack surface Storm-2603 exploited. Orgs running on-premises SharePoint servers should treat the following as a baseline.

  1. Patch internet-facing systems promptly, prioritizing critical infrastructure
  2. Establish a continuous process to identify, assess, and remediate vulnerabilities. Validate that updates have been applied
  3. Treat high-privilege identities as a primary attack surface: enforce strong identity controls and monitor privileged sign-ins
  4. Implement conditional access and identity protection to detect anomalous authentication activity early
  5. Deploy endpoint protection solutions across the entire environment
  6. Retain telemetry centrally (for example, in a SIEM) to support investigation, correlation, and timeline reconstruction
  7. Implement continuous monitoring and long-term log retention to surface malicious activity and enable complete investigations
  8. Audit and approve all remote management software, tunnelling utilities, and developer tooling in use. Monitor for atypical usage
  9. Develop predefined incident response playbooks, test them regularly, and ensure teams are prepared to act quickly
  10. Ensure compromised credentials, devices, and access paths can be rapidly isolated to limit lateral movement

Why the Attack Was Difficult to Detect

The attackers were methodical about the intrusion, knowing which tools security teams trust, which processes blend into normal administrative behavior, and how to stay hidden longer.

Abuse of Legitimate Tools

The attackers deliberately used software that would not inherently trigger security alerts. Velociraptor is a legitimate forensic tool; Cloudflare, Zoho Assist, and Visual Studio Code are all widely used applications. An antivirus engine scanning for known malicious software would have no obvious reason to flag these.

Active Disabling of Endpoint Protections

The report states the threat actor loaded a vulnerable driver (NSecKrnl.sys) to tamper with memory and disable endpoint protections. This means security tools were actively neutralized before they could detect subsequent activity.

Specific Camouflage

Looking at the activities uncovered by Microsoft, stated in their official report, the second actor hid malicious code inside legitimate processes. The trojan horse technique camouflaged the code, making it harder to distinguish from normal system activity.

Embedding Protection into SharePoint Repositories

The Storm-2603 attack shows that patches, identity controls, and endpoint detection remain essential, but they are not enough on their own.

Organizations should add an additional layer of file-level security for the content stored inside SharePoint repositories. While file security tools do not replace patches or endpoint defenses, they are nevertheless necessary; even if an attack breaches the perimeter, its blast radius stops short of the data itself.

Protect Your SharePoint Data from Further Attacks with MetaDefender™ Storage Security

Organizations can control is how well their data is protected once that happens.

That is the role MetaDefender™ Storage Security plays, bringing file security into SharePoint environments.

Real-time Scanning at the Storage Boundary

MetaDefender Storage Security supports real-time processing for SharePoint On-Premises. Every file that enters the environment is inspected before it can be accessed or shared, whether uploaded by a legitimate user or introduced through a compromised account.

Static and Dynamic Threat Detection

By combining Metascan™ Multiscanning (30+ anti-malware engines) with the Adaptive Sandbox technology (emulation-based dynamic analysis), organizations can expose evasion tactics that single-engine detection can miss.

For organizations using SharePoint, this means file-borne threats, including zero-days, are identified at the point of entry, not after the fact.

File Sanitization through the Deep CDR™ Technology

Not all files can be confirmed clean by detection engines, so MetaDefender Storage Security adds an additional security layer. Its Deep CDR™ Technology technology deconstructs files, removing potentially malicious content without disrupting the file's usability or the workflow it supports.

Data Loss Prevention

The Proactive DLP™ software detects, redacts, and anonymizes confidential information, preventing it from leaking in an attack. Even if threat actors establish can exfiltrate data over time, DLP at the storage layer limits the impact of a data leak. This layer doesn’t depend on whether the intrusion itself has been detected.

Audit Trail and On-Demand Scanning

MetaDefender Storage Security logs all scan activity and tags files, maintaining a record of what entered the storage environment and when.

In a post-incident investigation, security teams can run immediate on-demand scans across the full SharePoint environment to surface any malicious files at rest. This supports both remediation and the timeline reconstruction that incident response requires.

The Attacker’s Advantage

One attack doesn't tell you the whole story of a breach. Therefore, patching and moving on might not be the end of Storm-2603 exploits.

Two attackers, infiltrating through two different methods, in the same environment, means single-layer defense was never going to be enough. It also means there's no real way to estimate how many other systems, right now, are already breached, with an actor inside choosing to stay hidden. Patching closes the loophole that was exploited, but it doesn’t do anything for what may already lie dormant in your SharePoint libraries.

What’s in your control is ensuring your data inside is as hard to reach, move, and weaponize as possible.

Start at the file level with MetaDefender Storage Security and build from there.

Stay Up-to-Date With OPSWAT!

Sign up today to receive the latest company updates, stories, event info, and more.