Bastion Security

Exploiting Public Zero-days for Privilege Escalation

Starting in April 2026, a disgruntled security researcher going by aliases such as “Chaotic Eclipse” and “Nightmare Eclipse”.
Talk to an expert
August 5, 2026

Starting in April 2026, a disgruntled security researcher going by aliases such as “Chaotic Eclipse” and “Nightmare Eclipse” began releasing potent Windows privilege escalation vulnerabilities and proof-of-concept (PoC) exploit code to the public. These releases occurred outside of a typical vulnerability disclosure process, where a vendor would have the opportunity to construct a patch remediating the vulnerability prior to the vulnerability’s public release.

As these vulnerabilities were “zero-days”, this left a critical window of opportunity for threat actors to utilise these public exploits within their campaigns to degrade antivirus and escalate privileges after achieving initial access on a Windows workstation or server.

Zero-day vulnerabilities and other emerging threats feed into the cyber threat intelligence (CTI) utilised by Bastion during the planning of adversarial simulation exercises. Incorporating the latest CTI during the planning and execution phases ensures that exercises closely emulate the current tactics employed by cyber threat actors, as well as the wider threat landscape.

Exploiting YellowKey to Escalate Privileges and Impair Defences

One of these exploits, “YellowKey”, allows a BitLocker drive to be accessed through an Administrator command prompt within the Windows Recovery Environment (WinRE). Access to an unlocked BitLocker-encrypted drive within the WinRE would usually require the user to supply the BitLocker recovery key for the drive.

With access to the unlocked system drive, sensitive files which would usually be protected while operating Windows normally can be tampered with in WinRE. Bastion has utilised this on Adversarial Simulation engagements to drop command and control (C2) implants. We have successfully added a compromised account to the local Administrators group, and disabled EDR products by inserting Windows Defender Application Control (WDAC) policies preventing the EDR’s components from ever starting. For more information about blocking security products using WDAC, see Utilising WDAC to bypass Microsoft Defender.

As exploitation occurs within WinRE regular system logging does not occur making detection difficult. This highlights the importance of assuming breach and ensuring adequate detection of post-exploit attacker behaviour.

Obfuscating MiniPlasma

MiniPlasma exploits an incomplete fix for a previous CVE, CVE-2020-17103, exploiting the Windows Cloud Filter driver to escalate privileges by spawning a SYSTEM command prompt.

While detections were added to Windows Defender to detect the original PoC, Bastion was able to modify the exploit to execute undetected, achieving local privilege escalation on an initial foothold. By applying techniques such as string obfuscation and Import Address Table (IAT) hashing to hide suspicious imports, the original PoC could be modified to evade Defender’s static detection rules.  

Conclusion

Microsoft released fixes for YellowKey, MiniPlasma, and a third (GreenPlasma) in Microsoft’s June 2026 “Patch Tuesday” release.

“Nightmare Eclipse” has continued releasing new Windows exploits, their latest release at the time of writing this article being “LegacyHive” which could allow a threat actor with standard privileges to mount another user’s registry hive and escalate privileges.


Events

Latest events

Join Bastion experts for networking events, technical briefings, and hands-on workshops hosted throughout the year.
View all events
Cyber security news

Latest advisories

Stay ahead of emerging threats with our expert blog posts, research, and industry updates.
Access Token Exposure in URL Parameters in GridTime™ 3000 GNSS Time Server - (CVE-2026-12620)
During a security engagement, Leo Diamat discovered that the GridTime 3000 GNSS Time Server web application transmitted session access tokens via URL query string parameters on multiple endpoints.
Cross-Site Scripting (XSS) Vulnerability on Several Endpoints by Utilising Cross-Site Request Forgery (CSRF) in GridTime™ 3000 GNSS Time Server - (CVE-2026-12619)
During a security engagement, Leo Diamat discovered that multiple endpoints in the GridTime 3000 GNSS Time Server web application were vulnerable to reflected Cross-Site Scripting (XSS) via an unsanitised token parameter.
PHP-FPM (PHP Source) - Stored Cross-Site Scripting (XSS) (CVE-2026-6735)
During a security engagement, Conrad Draper discovered a stored XSS vulnerability in the PHP-FPM status endpoint which was due to a lack of input sanitisation of the request URI. This affects the request URI when displaying stored content.