FastUnlock: The Windows 11 Lock Screen Was Never a Security Boundary

The CyberSec Guru

FastUnlock Windows 11 Lock Screen Bypass

If you like this post, then please share it:

Buy me A Coffee!

Support The CyberSec Guru’s Mission

🔐 Fuel the cybersecurity crusade by buying me a coffee! Why your support matters: Zero paywalls: Keep the main content 100% free for learners worldwide.

“Your coffee keeps the servers running and the knowledge flowing in our fight against cybercrime.”☕ Support My Work

Buy Me a Coffee Button

On November 19, 2026, a researcher publishing as MSNightmare and Nightmare Eclipse pushed a repository called FastUnlock to GitHub. It contains no exploit code, just a README, an MIT license, and a demonstration video. Together they describe a way to make a locked, BitLocker-encrypted Windows 11 machine run a program of the attacker’s choosing without anyone signing in. ThreatLocker tested the technique the same day and confirmed it reaches code execution under a limited local account called WSIAccount, which the firm noted is “not an administrator account.” There is still no Microsoft patch, no CVE, and no public statement from the company, and no evidence the method has been used in a real attack. The Hacker News reported the disclosure on November 20.

The privilege level is modest, which is exactly why this is worth reading carefully instead of dismissing. FastUnlock does not break encryption, corrupt memory, or touch a remote service. It walks through a door Windows leaves open before authentication, and it exposes a gap between what people assume a locked, encrypted laptop guarantees and what the architecture actually promises. Following the keystrokes shows how far that gap reaches, especially once this first stage gets paired with a second unpatched bug the same researcher released nine days earlier.

How FastUnlock Gets a Shell Without Unlocking the Screen

The entry point is the “I forgot my PIN” link on the sign-in screen, which launches the Microsoft account recovery flow. From the recovery window, the instructions are to press Win+U for accessibility options and turn on Narrator, the built-in screen reader. The moment Narrator starts, it focuses on a background window it owns, one that, in the researcher’s words, “will be in the background but you would not be able to see it.” The attacker navigates that invisible window the way a blind user would: by listening to Narrator and stepping through its elements with the keyboard. Holding Caps Lock, the default Narrator modifier key, and using the arrow keys moves the focus to a “Next” button, and the target is a link labeled “Math reading settings.” Pressing Caps Lock+Enter activates it, and Windows Settings opens, still on the locked screen, still with no user signed in.

From Settings, the path to code execution runs through pages nobody associates with security. The researcher’s route goes through Bluetooth & devices, then AutoPlay, then Removable drive, then “Open folder to view files (File Explorer).” At that point the attacker inserts a USB stick. Narrator’s focus jumps to Explorer showing the mounted storage, and the same Caps Lock plus arrow keys, Enter, and Space combination launches a binary from the drive. The binary doesn’t run as the machine’s owner, since the owner never logged in. It runs as WsiAccount.

The root cause, per the researcher, is a redesign oversight. Microsoft gave Narrator a new interface and, in his phrasing, “forgot to make sure that the new UI does not run in the case it was running in the context of a lockscreen user.” The older guard that stops Narrator from spawning its window when it runs as SYSTEM is still in place. The new surface simply wasn’t added to the list of things that must not appear before authentication. This isn’t a memory-safety bug; it’s an access-control gap in a feature meant to work before sign-in, built from a sequence of legitimate keystrokes rather than a payload.

ThreatLocker’s independent test filled in the practical details. Its testers found the physical keyboard sometimes wouldn’t accept typing while Narrator was being navigated, but the on-screen keyboard worked when text input was needed. They used an external drive to deliver the payload and noted other delivery methods exist. The published video stops short of the final step; the researcher wrote he didn’t show full code execution “because I don’t really have a USB stick,” but that he’d done it enough times to know what happens next. His claim is that the whole process took five minutes, from discovery to breaching a fully encrypted Windows 11 26H2 machine running the November patch, KB5132061, released November 10, and that it was “extremely consistent in every machine I tested it in.” He hasn’t checked Windows 10 and says so plainly. There’s no official list of affected versions, so any machine meeting the preconditions should be treated as potentially exposed until Microsoft says otherwise.

What WSIAccount Is, and Why the Privilege Ceiling Still Matters

WSIAccount isn’t an ordinary user account, and understanding it is key to judging the severity here. Microsoft’s web sign-in documentation describes it as a system-managed local account used for Web sign-in and Temporary Access Pass, the passwordless and federated sign-in experiences. It’s created automatically when Web sign-in is enabled, never appears in the user selection list, is enabled each time the Web sign-in credential provider is used, and is disabled again after the user signs in. Its purpose is to host authentication-related UI in a constrained context before a real session exists. That’s the context FastUnlock’s payload inherits.

ThreatLocker’s point that this isn’t an administrator account is accurate and worth taking seriously. On its own, FastUnlock doesn’t hand over administrative control, doesn’t let the attacker reset the device owner’s password, and doesn’t expose the owner’s profile the way an authenticated session would. A limited token is a real ceiling. But treating that ceiling as the end of the story misreads how compromise usually proceeds. What the attacker has is interactive code execution on a machine whose drive is already unlocked, running under an identity that exists specifically to do pre-authentication work. That’s a foothold, and footholds are worth exactly as much as what can be chained onto them. The researcher calls FastUnlock a first stage, and he names the second.

SignDrop: The Privilege Escalation That Completes the Chain

On November 10, this month’s Patch Tuesday, the same researcher released SignDrop, a local privilege-escalation proof-of-concept with full code and a write-up. BleepingComputer reported it on November 12. It targets the Common Log File System, the kernel-mode logging subsystem behind clfs.sys, and specifically the authentication mitigation Microsoft shipped in October 2025 to defend against a long run of CLFS memory-corruption bugs. That mitigation attaches a hash-based message authentication code to each log file, computed with a system-unique cryptographic key stored in the registry and readable only by Administrators and SYSTEM. CLFS validates the code at open time, reading the whole file before parsing its structures, and refuses to open a log that fails the integrity check. A new system spends 90 days in learning mode, silently adding codes to existing logs, before moving to enforcement. The operating mode lives at HKLM\SYSTEM\CurrentControlSet\Services\CLFS\Authentication under the value Mode: 0 is enforced, 1 is learning, 2 means an administrator disabled it, 3 means the system disabled it.

The detail SignDrop abuses is the companion file. Alongside each binary log file (.blf) and its data containers, CLFS creates a .cnpf “patch file” holding recovery data for logs that weren’t cleanly closed. Microsoft’s documentation states this file “will be created with the same security attributes as the file it provides recovery information for.” That inheritance is the crack. As independent analyses describe it, a standard user creates a CLFS log and an associated container they control, makes the container empty so its .cnpf is missing, and plants an object manager symbolic link that redirects the .cnpf creation path somewhere of their choosing. They then coerce a privileged process into opening the log, with the published chain calling Windows Defender’s update interface, ServerMpUpdateEngineSignature. CLFS creates the .cnpf in the security context of the process that opened the log, so it’s owned by SYSTEM, but it inherits the original log’s discretionary access control list, which the standard user controls. The result is a SYSTEM-created file the attacker can write to, and that write gets chained into a SYSTEM shell.

📬 Stay Ahead of Cyber Threats

Get the latest cybersecurity news, critical vulnerabilities, threat intelligence, tutorials, and exclusive giveaways delivered straight to your inbox. No spam. Unsubscribe anytime.

The researcher’s framing is that Microsoft implemented the mitigation correctly, but “the logic allows abuse.” He makes a broader claim too: because the file-creation request happens inside the antimalware engine’s own trusted process, the technique can defeat any engine’s self-defense or tamper protection, since from the kernel driver’s perspective the request comes from a legitimate process. That claim reaches further than what’s actually been tested. He tested Windows 11 26H2, a 26H2 Experimental build, and a Server 2025 Experimental build, and estimates 80% of Windows installations are vulnerable, his estimate, not a measured figure. An independent assessment by 78researchlab, reported by The Hacker News, found the exploit works but isn’t reliable enough for operational use and is conspicuous: running it changes every User Account Control prompt to open both a standard-user shell and a SYSTEM shell. There’s no CVE and no fix, and the November updates didn’t address it. Disabling the mitigation is not a remedy; setting Mode to 2 or 3 reopens the memory-corruption exposure the feature exists to close.

Put the two together and the intended chain is clear: FastUnlock converts physical access to a locked machine into code execution as WsiAccount, and SignDrop converts a limited local session into SYSTEM. The researcher describes the combination as full control, “assuming your machine is vulnerable to SignDrop.” One caveat matters for accuracy: no public demonstration has yet chained FastUnlock into SignDrop end to end from the lock screen. ThreatLocker verified only the first stage, and the published SignDrop chain assumes a standard-user session with Defender enabled. Whether the WsiAccount context satisfies every precondition of the second stage hasn’t been shown publicly. But both primitives are public, both are unpatched, and the researcher presents them as two halves of one path. Defenders should assume chainability rather than wait for a polished demonstration.

How FastUnlock allows code execution before Windows 11 sign-in, the potential SignDrop privilege-escalation chain, and why BitLocker encryption itself remains intact
How FastUnlock allows code execution before Windows 11 sign-in, the potential SignDrop privilege-escalation chain, and why BitLocker encryption itself remains intact

Why This Isn’t Really a BitLocker Bypass

The most important nuance in this whole story is the one most headlines flatten. FastUnlock is being described as a BitLocker bypass, and in a loose sense it is, but it doesn’t attack the encryption. It doesn’t recover the volume master key, defeat the TPM’s sealing, or exploit a flaw in the cipher. It operates at a point in the boot sequence where BitLocker has already finished its job.

BitLocker encrypts volume data with a full volume encryption key, itself encrypted by a volume master key, which is in turn protected by a key protector: the TPM, a PIN, a startup key, or a recovery key. The default cipher is XTS-AES 128-bit, configurable to 256-bit. In the common TPM-only configuration, the TPM seals the volume master key against measurements of the boot path: Secure Boot state via PCR[7], and a PCR[11] value the boot manager guarantees to change before handing control to any loader, which stops a rogue operating system from unsealing the key. A memory-overwrite request mechanism defends against reset attacks. Microsoft’s countermeasures documentation describes TPM-only mode as one that “doesn’t require any interaction with the user to unlock and provide access to the drive.” If the measurements match, the drive unlocks silently and Windows boots straight to the sign-in screen. By the time that screen appears, the volume is mounted and readable, and the keys are already in memory. FastUnlock runs after all of that.

This is less a loophole in BitLocker than a misunderstanding of its scope. Microsoft’s security servicing criteria lists BitLocker as a feature whose stated goal is that “data that is encrypted on disk cannot be obtained when the device is turned off.” A running, locked machine isn’t turned off. The lock screen itself isn’t among the security boundaries Microsoft enumerates (kernel, process, user, session, network, and the virtualization boundaries), and features like User Account Control and AppLocker are explicitly classified as defense-in-depth, not serviced by default. The architecture is coherent; the mistake is treating a UI state as though it were a cryptographic one. Even Windows 11 25H2’s hardware-accelerated BitLocker, which offloads bulk cryptography to a dedicated engine and hardware-wraps the bulk keys, doesn’t change what’s exposed once the volume is unlocked.

Power state is where this gets operational. A sleeping machine keeps its keys in RAM and requires no reauthentication to reach data; a machine that’s on and locked has likewise already passed the pre-boot gate. Hibernation is different: the drive locks, and resuming requires the PIN or startup key again if multifactor protection is configured, though Microsoft notes this “doesn’t have an impact on TPM-only mode,” which stays transparent even on resume from hibernation. The countermeasures guidance is blunt about the determined attacker: configure a TPM+PIN protector, disable standby power management, and “shut down or hibernate the device before it leaves the control of an authorized user.” For an opportunistic attacker with limited physical access, Microsoft considers TPM-only sufficient. FastUnlock is a reminder that the line between those two attacker profiles is thinner than the default configuration assumes.

Which Windows 11 Machines Are Actually Exposed

FastUnlock isn’t universal. ThreatLocker’s test names three requirements: physical access to the machine, a Microsoft account reachable through the account recovery flow, and the ability to navigate Narrator with Caps Lock and the arrow keys. The account condition is more restrictive than it first looks. The recovery flow being abused is built for accounts linked to an online identity, on a consumer PC that means a Microsoft account, because the “forgot my PIN” path launches a web-based recovery process. A device configured exclusively with local accounts doesn’t present the same surface, and this is the researcher’s own description of the precondition, not something Microsoft has confirmed.

In practice the condition isn’t a high barrier. Windows 11’s setup has for years pushed users toward Microsoft accounts, and on many consumer devices the primary account is online-linked by default. Automatic device encryption, whose eligibility Windows 11 24H2 broadened by removing the DMA and HSTI/Modern Standby prerequisites, initializes with a clear key and only completes its protection, creating the TPM protector and uploading the recovery key, when an administrator signs in with a Microsoft account on a non-domain-joined device. A device using only local accounts from the start, Microsoft notes, “remains unprotected even though the data is encrypted.” The encryption and the account linkage are entangled by design.

On Windows 11 Home, Device Encryption offers no startup-PIN option at all, a point Privacy Affairs’ analysis makes explicit. On Pro, Enterprise, and Education, BitLocker’s default remains TPM-only unless an administrator has deliberately configured a TPM+PIN protector. So the exposed population is large: machines that are physically reachable, carry an online-linked account, and unlock their drives transparently. The population that also enforces a startup PIN is comparatively small.

Power state decides the rest. A laptop that’s asleep or sitting on the lock screen is exposed if it meets the account condition. A laptop shut down or hibernated with a TPM+PIN protector stops at the pre-boot PIN prompt before Windows and its accessibility tools ever load. ThreatLocker states this directly: a pre-boot PIN protects a machine that’s powered off or hibernated because it must be supplied before the drive unlocks. That’s an architectural reading rather than a Microsoft confirmation specific to FastUnlock, and there’s a precedent worth a note of caution. When the same researcher disclosed the YellowKey BitLocker bypass in May (CVE-2026-45585, CVSS 6.8, physical access), which abused Transactional NTFS replay in the Windows Recovery Environment to delete winpeshl.ini and serve a command prompt with the drive unlocked, Microsoft’s advisory FAQ stated that “if you are using TPM+PIN the vulnerability is not exploitable.” The researcher disputed that, claiming to hold back a variant that bypasses TPM+PIN. Microsoft shipped a mitigation on May 20, a script on May 21, and a fix on June 10. For FastUnlock the logic for a PIN holds up, since the lock screen never loads, but it hasn’t been vendor-confirmed, and YellowKey is a reminder that adding a PIN isn’t a universal guarantee against every bypass in this family.

How to Defend Against FastUnlock Before Microsoft’s December Patch

With no patch available, the useful work is about physical control and moving the authentication boundary earlier. Keep machines that can go missing out of the wrong hands, and treat shutdown discipline as a control rather than a courtesy: a sleeping laptop has its drive unlocked, and a startup PIN only helps a machine that hasn’t started yet. Shut down or hibernate devices before they leave someone’s control, and on managed fleets disable standby through the policies Microsoft’s countermeasures guide names: showing hibernate in the power options and disabling the S1 through S3 standby states on battery and plugged in.

On Pro, Enterprise, and Education, add a startup PIN. The Group Policy path is Computer Configuration > Administrative Templates > Windows Components > BitLocker Drive Encryption > Operating System Drives; enable “Require additional authentication at startup” and allow a startup PIN with TPM. Then, from an elevated prompt, Microsoft’s documented conversion is manage-bde -protectors -delete C: -type tpm followed by manage-bde -protectors -add C: -tpmandpin <PIN>. Make the PIN long and, ideally, alphanumeric. Enhanced PINs use the full keyboard character set, run 4 to 20 characters, and are gated by the “Allow enhanced PINs for startup” policy; the PIN is hashed with SHA-256 and the first 160 bits become the authorization data the TPM uses to seal the volume master key. The TPM’s anti-hammering protection, which exponentially increases lockout time after failed attempts, is what makes a short PIN viable at all.

That last point comes with a fresh caveat. On November 18, researchers at TU Berlin published “Hello Again,” showing the anti-hammering counter can be rolled back on affected platforms. Modern firmware TPMs persist their non-volatile state, including lockout counters, to SPI flash, and newer platforms delegate rollback protection to the flash chip via Replay Protected Monotonic Counters. Using a low-cost interposer, the team found Winbond flash leaks its HMAC key under voltage glitching and that AMD’s NV controller doesn’t verify RPMC increments, so counter updates can be suppressed unnoticed; AMD platforms without RPMC flash appear to lack NV rollback protection altogether. The effect reaches AMD, Intel, and Pluton fTPMs. Their end-to-end attack cuts the worst-case brute-force time for a default six-digit BitLocker PIN from 19 years to 20 days, and for a four-digit Windows Hello PIN from 69 days to 40 hours. AMD confirmed a firmware flaw, CVE-2026-43613, with patches rolling through February 2027; the Winbond issue has no committed fix. Microsoft rated the Hello case defense-in-depth and committed to raising the recommended minimum Hello PIN length to six digits.

The lesson for FastUnlock scales proportionately. This hardware attack needs an opened chassis and interposer skills, a different profile from someone holding a locked laptop for five minutes, but it means a short numeric PIN isn’t the wall it looks like. A long enhanced PIN, which Microsoft’s own guidance says can make brute force “nearly impossible,” and physical custody matter more than ever. For enterprises that need unattended reboots, BitLocker Network Unlock provides a wired-network second factor so TPM+PIN doesn’t strand remote machines.

On Home, where no startup PIN exists, physical control and shutdown discipline are the practical defenses. Converting every account to a local account removes the precondition the researcher names, but it’s a trade: it changes recovery-key escrow and sync, and Privacy Affairs rightly advises confirming the recovery key is saved at aka.ms/myrecoverykey before switching. There’s also a stopgap the researcher himself suggested on X: blocking Narrator with AppLocker or an access-control list on its executable, and a community App Control (WDAC) deny-rule script appeared on GitHub the same day. It’s third-party code, so review it before deploying, and weigh the cost: blocking Narrator takes the screen reader away from anyone who relies on it. Worth noting, Microsoft classifies AppLocker as defense-in-depth but WDAC as a serviced security feature, so a WDAC deny rule is the stronger of the two controls. Don’t disable CLFS authentication to dodge SignDrop; that trades one bug for a class of them.

For detection, SignDrop has published artifacts worth hunting: .cnpf files in user-writable directories owned by SYSTEM, object manager symbolic links pointing from the \CLFS namespace into user-writable locations, and the characteristic change in UAC prompt behavior. FastUnlock has no published indicators, so watching for process creation under WsiAccount, or Explorer and Settings launching before any interactive logon, is a reasonable hunting hypothesis rather than a confirmed signal. Microsoft’s next scheduled security updates are due December 8, 2026, and a fix could arrive then or out of band; watch for an advisory and apply it when it comes.

The Researcher Behind FastUnlock and the Disclosure Debate

FastUnlock can’t be separated from its author. The researcher is Abdelhamid Naceri, who has published as Chaotic Eclipse, INFINITE NIGHTMARE, MSNightmare, and Nightmare Eclipse, and earlier under the handle “halov.” He isn’t new to this surface: the Zero Day Initiative credits him with ZDI-21-1053, CVE-2021-26431, a Windows lock-screen improper-access-control authentication bypass with a physical-access CVSS of 6.8, reported in June 2021 and patched that September. He also went through coordinated disclosure for a Microsoft zero-day in 2020 and a Kaspersky flaw in 2021 via ZDI. The FastUnlock README itself references ZDI-21-1053 as a prior lock-screen bypass he found, noting that this one is “extremely convenient to use” by comparison.

Security Affairs reports that Naceri previously worked with Microsoft in the UK and Germany, and BankInfoSecurity’s account of his identity reveal in September 2026 describes a termination dispute: a 2024 escalation over sharing vulnerability information, a termination letter dated March 2025, and a German labor court that upheld the dismissal, according to his account. Since April 2026 he has published a run of Windows zero-days without warning Microsoft first: BlueHammer, RedSun, UnDefend, YellowKey, GreenPlasma, MiniPlasma, RoguePlanet, GreatXML, LegacyHive, ShieldBreak, ShieldCrash, and BigDiskBuster, alongside exploits aimed at Kaspersky (HardBreacher), Avast (PrettyPrague), CrowdStrike Falcon (FalconFlank), and Nvidia (GreenSection). He has settled into announcing on Patch Tuesdays, leaving little runway for mitigations; after FastUnlock he posted, “aww too bad almost 4 weeks till next patch tue.” Microsoft threatened legal action in May against “malicious activity causing real harm” and later walked the statement back after community backlash. Some of his disclosures have been fixed, including ShieldBreak, RoguePlanet, YellowKey, GreenPlasma, and MiniPlasma, and Security Affairs notes that some earlier ones were later exploited in the wild.

The ethics here are genuinely contested and worth stating plainly rather than dramatically. Publishing unpatched bypasses and privilege escalations gives defenders test cases and attackers blueprints, and the balance depends on whether a vendor fix is in progress and whether the most damaging step is included. FastUnlock’s README omits the final USB execution step, but ThreatLocker supplied it; SignDrop ships full code. No vendor fix has been confirmed for either, and the December window is real. Defenders have to operate inside that uncomfortable middle ground rather than wait for it to resolve.

What FastUnlock Reveals About Windows Lock Screen Security

The pattern here is older than this researcher. Accessibility and recovery features that must work before authentication have repeatedly produced unintended code-execution paths on the lock screen, and Microsoft has serviced lock-screen bypasses before, including Naceri’s own 2021 CVE. FastUnlock is novel in its mechanism, a redesigned Narrator interface running in the lock-screen account context, but not in its category. The fix, when it comes, will almost certainly address Narrator’s behavior in that context rather than BitLocker, because BitLocker is not what failed.

What failed is an assumption. A locked screen is a UI state, not one of the security boundaries Microsoft defines, and BitLocker’s serviced promise is explicitly about the powered-off device. A running, locked, TPM-only-encrypted laptop is a live host whose drive is already readable, and FastUnlock shows that five minutes and a USB stick are enough to make it run someone else’s code. Until a patch lands, the posture that matches the architecture is the one Microsoft’s own countermeasures guidance already describes: keep the machine in custody, move authentication to pre-boot with a long enhanced PIN where the edition allows it, and shut down or hibernate before it leaves controlled hands. The lock screen was never the boundary. It’s time the default configuration stopped pretending it was.

Buy me A Coffee!

Support The CyberSec Guru’s Mission

🔐 Fuel the cybersecurity crusade by buying me a coffee! Your contribution powers free tutorials, hands-on labs, and security resources.

Why your support matters:
  • Writeup Access: Get complete writeup access within 12 hours
  • Zero paywalls: Keep the main content 100% free for learners worldwide

Perks for one-time supporters:
☕️ $5: Shoutout in Buy Me a Coffee
🛡️ $8: Fast-track Access to Live Webinars
💻 $10: Vote on future tutorial topics + exclusive AMA access

“Your coffee keeps the servers running and the knowledge flowing in our fight against cybercrime.”☕ Support My Work

Buy Me a Coffee Button

If you like this post, then please share it:

Exploits

Discover more from The CyberSec Guru

Subscribe to get the latest posts sent to your email!

Leave a Comment

This site uses Akismet to reduce spam. Learn how your comment data is processed.