A recently disclosed privilege escalation vulnerability in Microsoft Defender has been exploited in the wild as a zero-day using publicly available proof-of-concept (PoC), Huntress warns.
Patched on April 14, the issue is tracked as CVE-2026-33825 (CVSS score of 7.8). Microsoft describes it as an elevation of privilege bug rooted in insufficient granularity of access control.
The CVE was publicly disclosed on April 2 by a disgruntled researcher known as Chaotic Eclipse and Nightmare-Eclipse, who warned it was a race condition leading to full System privileges.
The researcher named the flaw BlueHammer and published PoC exploit code to their GitHub repository. Interest in the exploit surged fast, fueled by a fork that fixed some bugs in the researcher’s implementation and included documentation and instructions.
BlueHammer is a time-of-check to time-of-use (TOCTOU) in Defender’s signature update mechanism that allows an attacker with low privileges to gain System permissions.
The first attacks leveraging the public PoC were seen on April 10, with additional activity observed on April 16, cybersecurity firm Huntress warns.
“Huntress identified suspicious FortiGate SSL VPN access tied to the compromised environment, including a source IP geolocated to Russia, with additional suspicious infrastructure observed in other regions,” the company says.
The attacks leveraged all three techniques that Chaotic Eclipse published, namely BlueHammer, RedSun, and UnDefend.
BlueHammer relies on operation locks (oplocks) to suspend Defender’s operation and on triggering a signature update to trick Defender into copying the Security Account Manager (SAM) database to its output directory.
BlueHammer then parses the SAM hive, decrypts users’ NT hashes, temporarily changes all user passwords to a new one, and uses the new password to generate admin sessions that can be used to gain System permissions.
RedSun works the same, but relies on rewriting critical system files to achieve System privileges. It tricks Defender into attempting to restore a non-existent ‘malicious file’ to place a copy of itself in the System32 directory, and then spawns a shell with System permissions.
UnDefend kills Defender by locking definition files. For that, it monitors for changes to the definition updates and Microsoft’s Malicious Software Removal Tool folders to lock new files before Defender can use them, and locks backup definition files immediately after Defender’s startup.
“One of the clearest patterns Huntress observed was the use of user-writable directories for staging and execution. In the most recent occurrence, binaries were staged from a low-privilege user’s Pictures folder and short two-letter subfolders under Downloads,” Huntress says.
The cybersecurity firm says the attackers accessed the target environment through an SSL VPN connection to a FortiGate firewall. The hackers were not familiar with how the Defender exploits worked and were unsuccessful in their attempts, but did perform hands-on keyboard reconnaissance operations.
On Wednesday, the US cybersecurity agency CISA added CVE-2026-33825 to its Known Exploited Vulnerabilities (KEV) catalog, urging federal agencies to patch it by May 6.
Related: Organizations Warned of Exploited Cisco, Kentico, Zimbra Vulnerabilities
Related: Recent Apache ActiveMQ Vulnerability Exploited in the Wild
Related: Exploited Vulnerability Exposes Nginx Servers to Hacking