The proof-of-concept went up on GitHub on Nov. 10, Patch Tuesday, the day Microsoft ships its monthly fixes. It does not attack a new bug. It attacks a fix.

The exploit is called SignDrop, and it lets an ordinary Windows account become NT AUTHORITY\SYSTEM, the account that runs the kernel and the antivirus engine and sits a rung above Administrator. What makes it worth a close look is where it lands: inside the authentication layer Microsoft bolted onto the Windows Common Log File System driver last year for the express purpose of ending privilege-escalation attacks against that driver. The defense is the way in.

CLFS has been the single most reliable route from a foothold to full control of a Windows machine for ransomware crews for four straight years. Microsoft tried to close that route with cryptography. A researcher it fired has now turned the cryptography into the seventh candidate for the same list. As of Nov. 16 there is no CVE number and no patch.

A decade of one driver

The Common Log File System is a general-purpose logging engine built into Windows. Programs hand it records; it packs them into log files and plays them back. The reason it keeps turning up in attacks is structural. CLFS parses those log files inside the kernel, in clfs.sys, so a flaw in the parsing code runs with kernel privileges. Find a way to feed it a malformed log and you are executing with the operating system’s own rights.

A cream shield with a coral stripe contains an open door and a large dark blue key against a bright blue background.

Microsoft has patched at least 32 local privilege-escalation holes in the driver since 2018, according to Kaspersky’s Securelist, which has tracked the driver as a favorite of financially motivated groups. Six of those were caught being exploited before any patch existed. The count by year reads two, two, one, one — 2022, 2023, 2024, 2025 — never zero across four years of patching.

The first pair, CVE-2022-24521 and CVE-2022-37969, and a 2023 pair, CVE-2023-23376 and CVE-2023-28252, sit in the Known Exploited Vulnerabilities catalog kept by the Cybersecurity and Infrastructure Security Agency. Kaspersky tied the Nokoyawa ransomware group to several of them and, after taking the exploits apart, concluded a single developer had built them. CVE-2024-49138 followed in December 2024.

The fix

The sixth, CVE-2025-29824, is the one that explains the mitigation SignDrop defeats. Microsoft disclosed it on April 8, 2025, the same day it shipped the patch, and attributed the attacks to a group it tracks as Storm-2460. As The Hacker News reported, the group used a backdoor called PipeMagic to run the CLFS exploit in memory, then dumped credentials and dropped a ransom note linked to the RansomEXX brand. Targets were few and scattered: information-technology and real-estate firms in the United States, a software company in Spain, a retailer in Saudi Arabia, a financial firm in Venezuela. It was the first confirmed case of a CLFS zero-day used to deliver ransomware rather than for espionage.

A dark silhouette climbs ladder rungs on a tall cream wall split by a jagged red crack against a vivid blue background.

That record is why Microsoft stopped patching individual bugs and tried to wall off the class. Starting in 2024 it experimented with a mitigation that gives each log file a cryptographic authentication code, an HMAC built from the file’s contents and a key unique to each machine, and stores it in a companion file with the extension .cnpf. Before the driver trusts a log, it checks the code. A log tampered with outside the driver’s control fails the check and will not open. The mitigation shipped in the October 2025 Windows update and is active in the 26H2 release.

On paper it is a clean idea: sign the logs, and an attacker can no longer smuggle a doctored one past the kernel. The flaw is not in the cryptography.

How the fix broke

The weakness sits in the bookkeeping around the signature, not the signature itself. When CLFS opens a log container that is empty, it signs the container and writes the .cnpf file — whether or not a file already sits at that path, and regardless of who controls the path. The researcher who published SignDrop put it plainly, writing that Microsoft’s work was correct but the surrounding logic invited abuse.

The chain pulls in Windows Defender. A standard user can point a CLFS log at a path they control, then call Defender’s signature-update interface so that Defender, running as SYSTEM, is the process that opens the log and makes the driver write the authentication file. Because of the redirect, that privileged write lands where the attacker chose. The written file is then replaced with a malicious library, a privileged process loads it, and the shell that spawns runs as SYSTEM. The same logic, the researcher said, lets a local process make a security product’s own kernel driver write a file on its behalf, because from the driver’s view the request is coming from its own trusted process.

What makes it funnier is the fact that it can be used against any antimalware engine to bypass their self defense (or tamper protection) mechanism as you can trigger arbitrary file creation in the context of their own antimalware process so from the perspective of their kernel drivers, the file creation request is coming from their own legit process so it passes, this makes this vulnerability unique as it can be used to turn off any antivirus/EDR.

Nightmare Eclipse, the researcher who published SignDrop, writing in the exploit’s release notes, securityaffairs.com

It is a local attack. Someone, or something, already has to be running code on the machine. SignDrop is what turns that beachhead into total control, the step ransomware needs after it gets in and before it can switch off defenses and encrypt.

Who can reproduce it

The researcher says SignDrop works on Windows 11 26H2, Windows 11 26H2 Experimental and Windows Server 2025 Experimental, and claims it should run on about 80 percent of the world’s Windows installations. The qualifier is in his own phrasing: you are vulnerable mainly if the install is old enough.

You’re generally vulnerable if your Windows installation is older than 90 days.

Nightmare Eclipse, in notes reported by SecurityWeek, securityweek.com

Independent testing confirms the exploit but not the reach. Will Dormann, a vulnerability analyst, said in an interview with Ars Technica that he reproduced it and it works as described, but not everywhere: it seems to need 26H2 on an install more than 90 days old, so a fresh 26H2 machine brought fully up to date would not do. For a system to be exposed, he said, the registry value under HKLM\SYSTEM\CurrentControlSet\Services\CLFS\Authentication\Mode has to be 0 or 1 — learning or enforcement mode. Of the many machines he checked, one met that condition.

Cyderes, a cybersecurity services firm, compiled and ran the code and got a SYSTEM shell, with no sign of use outside the lab. So the gap is not whether SignDrop works — two outside parties say it does — but how many real machines sit in its window. The researcher’s 80 percent and Dormann’s single confirmed box have not been reconciled, and the number that would settle it is one Microsoft has not published.

A magnifying glass highlights a cream rectangular block with a glowing red opening among rows of similar blocks on a teal background.

The researcher Microsoft fired

Nightmare Eclipse is Abdelhamid Naceri, a former Microsoft security researcher based in Germany who had been credited by Trend Micro’s Zero Day Initiative since 2020. His identity became public in September, in a series of posts he flagged “STORY TIME” on an X account that has since been suspended. By his own account, reported by BankInfoSecurity, Tom Gallagher, a vice president at the Microsoft Security Response Center, told him on Sept. 20, 2024, that he was being terminated for sharing vulnerability information with outside parties; his termination letter was dated March 2025.

Naceri sued in a Cologne labor court for unfair dismissal, saying Microsoft never disclosed details of the breach it blamed on him. He says the company offered settlements that climbed from 20,000 euros to 27,000, then to 55,000 euros plus a year’s pay, that he refused them, and that the court upheld the firing anyway, leaving him more than $200,000 in legal fees and, he says, in treatment at a psychiatric hospital.

One detail argues for caution with the rest. Naceri has admitted that an earlier claim of his — that Microsoft had sued him — was invented.

MS never sued, I lied.

Abdelhamid Naceri, retracting an earlier claim on X, bankinfosecurity.com

That admission is reason to treat his account, which Microsoft has not confirmed in detail, as one side of a dispute. SignDrop is his thirteenth public zero-day since April, after a run that includes BlueHammer, RedSun, UnDefend, YellowKey, GreenPlasma, MiniPlasma, RoguePlanet, GreatXML, LegacyHive, ShieldBreak, ShieldCrash and BigDiskBuster. Microsoft has patched some, including ShieldBreak, RoguePlanet, YellowKey, GreenPlasma and MiniPlasma, and left others unfixed.

Detection, not a patch

The run pushed Microsoft into print. On May 27 its response center published a post calling uncoordinated disclosure that puts proof-of-concept code into the wrong hands “never justifiable,” and adding a line that read to much of the security industry as a threat against a person rather than against organized crime.

Our Digital Crimes Unit will continue bringing cases against these actors and those that enable their criminal activity – coordinating as needed with law enforcement around the world.

Microsoft Security Response Center, in a May 27 blog post, microsoft.com

The backlash was fast. Microsoft walked it back within days, telling Windows Central it had “no intention to pursue action” against individuals doing or publishing security research, and quietly dropped the phrase “responsible disclosure” from its language. On SignDrop itself, Microsoft told SecurityWeek it is investigating and will patch affected products, and that Defender already catches the exploit.

Microsoft Defender detects and blocks exploitation of this vulnerability and alerts on the associated privilege-escalation behavior.

A Microsoft spokesperson, in a statement to SecurityWeek, securityweek.com

It separately told Ars Technica it has seen no exploitation in the wild. Neither statement says which Defender versions carry the detection or how completely they block the chain — and the chain runs through Defender itself. It is the same shape of answer Microsoft gave this month to a Secure Boot bypass thrown at one of its mitigations in public: a statement about detection in place of a fix.

What comes next

For now the practical check is the registry. A machine is in scope if the Mode value under the CLFS Authentication key is 0 or 1; administrators can read it directly, and can see which Windows build they are on with the winver command. Because Microsoft rolls out 26H2 in stages, it has not said — and may not yet know — how many machines meet SignDrop’s two conditions, 26H2 installed and more than 90 days old.

That number is the one that matters. The CLFS record shows how long the window between a working exploit and a patch has stayed open before: five of the last six exploited bugs in the driver were used in attacks before a fix shipped. Whether SignDrop becomes the seventh depends on a CVE and a patch from Microsoft, and on whether a ransomware operator judges the conditions common enough to weaponize first. Microsoft says a patch is coming. It has not said when.