SignDrop Windows Zero-Day: SYSTEM Escalation via CLFS and Defender
A public proof-of-concept now exists for an unpatched Windows local privilege escalation (LPE) called SignDrop. It abuses a mitigation Microsoft added in October 2025 to harden the Common Log File System (CLFS) driver. It needs local code execution first, so it is not an entry vector. It is the kind of tool an intruder uses after a phish or a stolen low-privilege account to reach SYSTEM. This article separates what is confirmed from what is still the researcher's claim.
What happened
Security researcher Abdelhamid Naceri, who publishes as Nightmare Eclipse (GitHub handle reported as MSNightmare), released the SignDrop exploit. Reporting from IT-Connect and Born's Tech Blog dates the release to November 10, 2026, which was Patch Tuesday. BleepingComputer, SecurityWeek, Ars Technica and several others covered it over the following days. Naceri later added a more detailed technical write-up, according to Born's Tech Blog.
The disclosure follows a pattern. BleepingComputer's coverage says Naceri has released a string of Windows zero-days this year, including BlueHammer, RedSun, UnDefend, YellowKey, GreenPlasma, MiniPlasma, RoguePlanet, ShieldBreak and BigDiskBuster. Microsoft has fixed some of them, such as ShieldBreak, RoguePlanet, YellowKey, GreenPlasma and MiniPlasma. Others still have no official patch.
Status of the facts as of publication:
- CVE: none assigned in the public reporting reviewed.
- Vendor patch: none. Born's Tech Blog says Microsoft had not confirmed or fixed the flaw, and that the issue had not previously been reported to Microsoft.
- Exploitation in the wild: not confirmed. Cyderes compiled and tested the PoC and found no evidence of active exploitation, per SecurityWeek.
- Independent validation: security researcher Will Dormann told Ars Technica he reproduced the exploit and that it works as described, though not on every Windows 11 build.
- CISA/CERT advisory: public reporting does not show one, and the flaw is not known to be in the KEV catalog.
- Victim count and threat-actor attribution: none reported.
Who is affected
Naceri says he tested SignDrop on Windows 11 26H2, Windows 11 26H2 Experimental, and Windows Server 2025 Experimental. He estimates that about 80% of Windows installations are vulnerable and says a machine is "generally vulnerable" if the installation is older than 90 days.
Treat those figures with caution:
- The 80% number is the researcher's estimate. Born's Tech Blog notes it is not verified with field data.
- Dormann's reproduction suggests a fresh install of the affected release with current updates may not be exploitable. Ars Technica relays his view that the target likely needs a machine that has been running that release for more than 90 days. Public reporting does not explain why the age of the installation matters.
- Reporting also says Microsoft Defender must be active, because the exploit uses Defender as the privileged process. Hosts running a third-party AV with Defender fully disabled may behave differently. That has not been publicly tested.
- Public reporting does not give a full matrix of vulnerable builds, so older Windows 10 or Server 2022 behavior is unconfirmed. Do not assume those systems are safe. Test them.
No sector or region targeting has been reported. Any environment with standard-user access on affected builds is in scope. Terminal servers, VDI pools, shared workstations and developer machines carry the most risk, because low-privileged users and untrusted code already run there.
How the attack works
The following is based on the researcher's description as relayed by BleepingComputer, SecurityWeek and others. It is not on-system analysis.
Background: the CLFS mitigation
CLFS is a Windows logging subsystem used from user mode and kernel mode, and its driver (clfs.sys) has a long history of memory-corruption bugs exploited for privilege escalation. In October 2025 Microsoft added an authentication mitigation. Log files are tied to a machine-specific authentication code stored in a companion file with the .cnpf extension. This stops attackers from feeding the driver tampered log files. Naceri says CLFS exploitation dropped sharply afterward.
The logic flaw
Naceri says the bug lies in how CLFS creates the .cnpf authentication file when a new log file is created. If the file does not exist, CLFS creates it, and it gives the new file the access permissions of the original file. In his words, Microsoft implemented everything correctly but the logic allows abuse.
Exploit chain (as reported)
- A standard user prepares a CLFS log file. The user uses Object Manager symbolic links so that the expected
.cnpfpath redirects to an attacker-chosen location. - The user calls Windows Defender's signature-update RPC interface. That forces the Defender engine, which runs as SYSTEM, to open the prepared file.
- Because the authentication file is absent, CLFS creates it in the Defender process context with SYSTEM privileges, at the attacker-controlled path.
- The PoC reopens that file and writes a payload into it, and a SYSTEM shell spawns.
The result is an arbitrary file-creation primitive in a SYSTEM-privileged antimalware process. Windows Report and other outlets say the same primitive could be used to tamper with Defender and bypass AV/EDR tamper protection. Naceri says this is what sets SignDrop apart. These are claims about capability, and no public analysis demonstrates the impact against specific EDR products.
MITRE ATT&CK mapping
- T1068 – Exploitation for Privilege Escalation: the core behavior, going from standard user to SYSTEM through a flaw in a Windows component.
- T1562.001 – Impair Defenses: Disable or Modify Tools: applies only if an operator uses the SYSTEM-level primitive to interfere with Defender or other security tools. That follow-on use is claimed but not demonstrated in public reporting.
Initial access, persistence and other techniques depend on the operator, not on SignDrop itself.
IOCs and identifiers
- CVE: none assigned.
- Hashes, domains, IPs: none published. This is a source-code PoC, so compiled binaries will vary by builder.
- Behavioral artifacts:
.cnpffiles appearing in unexpected locations. CLFS-related registry key:HKLM\SYSTEM\CurrentControlSet\Services\CLFS\Authentication\Mode. A SYSTEM shell spawned from a low-privileged session. - Source: the PoC is on the researcher's GitHub, per Windows Report. The repository is not linked here on purpose.
Detection guidance for defenders
No vendor detection content or official signatures have been reported. Detection therefore has to rest on behavior. These are the best candidates, and each needs baselining in your own environment.
What to look for
- .cnpf file creation outside expected paths. Public reporting does not state the normal locations of legitimate
.cnpffiles, so baseline them first. Then alert on creations outside those paths, or by unexpected processes. Pay special attention toMsMpEng.exe(Defender) creating them. - Changes to the CLFS Authentication Mode value. Naceri says disabling the mitigation reopens memory-corruption exploitation, so any change is high-signal.
- Privilege jumps. An interactive low-privileged session followed shortly by a SYSTEM
cmd.exeorpowershell.exe, particularly when the parent is not a normal service host. - Defender anomalies. Unusual signature-update RPC activity from a user process, Defender service errors, or tamper-protection events after a user logon.
- Unknown executables run from user-writable paths (Downloads, Temp, AppData) shortly before any of the above. These are common PoC staging locations.
Example hunting queries (Microsoft Defender XDR / KQL)
Query 1 finds .cnpf files created or modified, so you can baseline the legitimate pattern and spot odd initiating processes.
DeviceFileEvents
| where Timestamp > ago(14d)
| where FileName endswith ".cnpf"
| where ActionType in ("FileCreated","FileModified","FileRenamed")
| summarize Events=count(), Hosts=dcount(DeviceName), FirstSeen=min(Timestamp)
by InitiatingProcessFileName, InitiatingProcessAccountName, FolderPath
| order by Events asc
Rare combinations of initiating process and folder at the top are the ones to investigate. Anything created by MsMpEng.exe outside the baseline deserves priority.
Query 2 finds changes to the CLFS authentication setting.
DeviceRegistryEvents
| where Timestamp > ago(30d)
| where RegistryKey has @"Services\CLFS\Authentication"
| where ActionType in ("RegistryValueSet","RegistryKeyCreated","RegistryValueDeleted")
| project Timestamp, DeviceName, InitiatingProcessFileName,
InitiatingProcessAccountName, RegistryValueName, RegistryValueData
Any hit outside a documented change window is suspect.
Query 3 looks for SYSTEM shells launched by processes that were running as a standard user.
DeviceProcessEvents
| where Timestamp > ago(7d)
| where FileName in~ ("cmd.exe","powershell.exe","pwsh.exe")
| where AccountName in~ ("system") or AccountSid == "S-1-5-18"
| where InitiatingProcessIntegrityLevel in ("Medium","Low")
| project Timestamp, DeviceName, FileName, ProcessCommandLine,
InitiatingProcessFileName, InitiatingProcessFolderPath, InitiatingProcessAccountName
This catches a SYSTEM shell whose parent ran at medium or low integrity, a classic LPE signature. Expect some noise from admin tooling, so allowlist known parents. Field availability varies by tenant and sensor version.
None of these queries has been validated against a live SignDrop run in the public record, so tune them in a lab with the PoC before relying on them for alerting. Because the primitive may be used to weaken Defender, also alert when a host stops reporting telemetry or when tamper-protection status changes.
Mitigation and response
There is no patch, and the researcher says the only option is to wait for one. Disabling the CLFS mitigation is not a good workaround, because it exposes hosts to memory-corruption CLFS exploits. These steps reduce risk in priority order.
- Do not change the CLFS Authentication Mode value. Make this a standing instruction to admins, and alert on any change using Query 2.
- Cut off the precondition: local code execution. SignDrop only matters after an attacker has a foothold. Tighten application control (WDAC or AppLocker) on user-writable paths, review standard-user rights on servers, and push stronger email and web filtering. These measures break the chain before it reaches this exploit.
- Prioritize high-exposure hosts. This means RDS/VDI, jump hosts, developer endpoints and shared kiosks on Windows 11 26H2 and Server 2025. Remove unnecessary interactive logon rights there.
- Deploy the detections above and run the PoC in an isolated lab VM on a representative build. That tells you whether your fleet is exposed and whether your EDR sees the behavior. Do not run it on production machines.
- Protect your security stack. Confirm Defender tamper protection is on and monitored. Alert on sensor health gaps. If you run a third-party EDR, ask the vendor whether it has coverage for SignDrop behavior.
- Prepare to patch fast. Watch the Microsoft Security Update Guide and the next Patch Tuesday. When a fix ships, test and deploy to the high-exposure group first. Check for a CVE assignment and any CISA KEV listing, and pull the fix into your emergency-change path if exploitation is reported.
- If you find suspected exploitation: isolate the host, capture memory and the relevant
.cnpfand CLFS log files, then review the originating user account and how it got a foothold. Treat any SYSTEM-level compromise as full host compromise, so rotate credentials cached on that machine and rebuild it rather than clean it.
FAQ
Is SignDrop being exploited in the wild?
Not as far as public reporting shows. Cyderes found no evidence of exploitation after testing the PoC. With public code available, that could change quickly, so detection work should not wait.
Does SignDrop let attackers break in remotely?
No. It is a local privilege escalation, so an attacker needs to run code on the machine first. The risk is in turning a low-privilege foothold into SYSTEM and potentially disabling security tools.
Is there a CVE or a Microsoft fix?
No CVE or patch has been reported. Born's Tech Blog says Microsoft has not confirmed or fixed the issue. Microsoft has patched several earlier Naceri disclosures, but timing for this one is unknown.
Bottom line
SignDrop is a serious post-compromise tool, not an emergency for every desktop. It is dangerous because it is public, works through a trusted Defender process, and could help an attacker neutralize endpoint protection. Until Microsoft ships a fix, the best defense is to limit what low-privileged users can run and to alert on the file, registry and process behaviors above. The researcher's 80% exposure claim is unverified, so test your own builds rather than assuming either way.
Original reporting: BleepingComputer
Protect yourself with tools recommended by cybersecurity professionals:
The tools below are independently selected based on security audits, transparency, and real-world effectiveness.