EthanZer0/FaceLogin

简体中文 · 한국어 · English

A camera-based face-recognition unlock system built on the Windows Credential Provider framework.
A "Face Login" tile integrates into the lock screen — look at the camera and unlock. Supports both local accounts and Microsoft online accounts (MSA).

---

Features

---

System Architecture

flowchart TB
    subgraph LockScreen["Windows Lock Screen"]
        LogonUI["LogonUI.exe"]
        CP["FaceLogin<br/>CredentialProvider.dll"]
    end

    subgraph Service["Face Authentication Service"]
        Svc["FaceLoginService.exe"]
    end

    subgraph Enrollment["Enrollment Console"]
        Console["FaceLoginConsole.exe"]
    end

    subgraph Storage["Data Storage"]
        direction LR
        UsersDB[("data/<br/>users.dat")] ~~~ Config[("data/<br/>config.json")] ~~~ Models[("models/<br/>ONNX + landmarks")] ~~~ Logs[("log/<br/>log files")]
    end

    LogonUI -->|"COM interface"| CP
    CP -->|"Named pipe"| Svc
    Svc -->|"Credential handback"| CP
    Console -->|"RELOAD_DB / GET_LOGS"| Svc
    Console --> Storage
    Svc --> Storage

    style LockScreen fill:#eff6ff,stroke:#3b82f6
    style Service fill:#fefce8,stroke:#eab308
    style Enrollment fill:#f0fdf4,stroke:#22c55e
    style Storage fill:#fdf2f8,stroke:#ec4899

All cross-process communication goes through the DACL-protected named pipe \\.\pipe\FaceLoginPipe.

---

Star History

The chart is updated daily by the independent project StarHistory's GitHub Actions. Data and rendering are fully self-hosted with no third-party dependencies.

---

Quick Start

Step 1: Install

Download FaceLoginSetup.exe from Releases, run it, choose an install directory, and click Install.

Step 2: Enroll your face

Run FaceLoginConsole.exe as administrator. New installations default to no liveness check (none); if Blink or Anti-Spoof is enabled in Settings, follow its prompt. Then enter your password and click Save & Enroll.

Step 3: Unlock

Press Win + L to lock the screen, select the Face Login tile, and look at the camera. During ordinary unlock, selecting the tile arms input detection; the next keyboard key or mouse-button press starts recognition. Mouse movement alone is ignored. Recognition status is shown in a central overlay on the secure desktop. Cold boot starts recognition automatically by default, and can be configured to wait for a key or mouse button.

If the pose is outside the accepted range, the lock screen shows a localized instruction for turning left/right, looking up/down, or tilting your head. Recognition resumes when the pose is valid.

Uninstall

Run the installer and switch to the Uninstall tab, or launch Uninstall.exe from the installation directory. The uninstaller stops/removes the service, unregisters the Credential Provider, and removes the installation directory and FaceLogin data. Back up any data you want to keep before uninstalling.

---

System Requirements

Requirement	Details
OS	Windows 10 21H2+ / Windows 11 (x64)
Camera	USB or built-in, 1280×720 supported
Runtime	WebView2 (built into Windows 11, auto-installed on Windows 10)
Privileges	Administrator (required for installation and enrollment)
Disk space	Installer ~100 MB; installed programs, runtimes, and models ~110 MB, plus user data

---

Security Design

Layer	Measure
Process communication	Named pipe DACL: SYSTEM + Administrators only, remote access denied
Credential storage	DPAPI CRYPTPROTECT_LOCAL_MACHINE machine-scope encryption
Memory protection	Password wiped with SecureZeroMemory immediately after use
Liveness and pose	Liveness is configurable; lock-screen recognition applies a MobileNetV2 head-pose gate before matching
Camera handling	Media Foundation is preferred with DirectShow fallback; the current release has no software exposure normalization, so camera auto-exposure remains under the capture backend
Match security	Euclidean distance threshold + best/second-best match ratio dual check
Build hardening	ASLR, DEP, CFG, 64-bit high-entropy address randomization

---

Project Structure

FaceLogin/
├── common/                 # Shared library (logging, IPC protocol, DPAPI, account info, config)
├── credential_provider/    # Windows Credential Provider COM DLL
├── face_service/           # Face recognition Windows service
├── enrollment_app/         # Face enrollment console (WebView2 GUI)
├── installer/              # Go Wails graphical installer
├── locales/                # Standalone language packs (zh-CN / ko-KR / en-US)
├── scripts/                # Build scripts and locale-pack consistency checks
└── assets/                 # Icon resources

See DEVELOPMENT.md for the detailed technical documentation.

---

Building from Source

Only needed when compiling yourself.

Prerequisites

• Visual Studio 2022 (with the C++ workload)
• vcpkg — dlib (image utility library: matrix/rectangle/transform), onnxruntime
• Go 1.25+ + Wails v2 (installer only)
• CMake 3.20+

C++ Components

# vcpkg dependencies (recognition/detection/landmarks/pose/anti-spoof are all ONNX; dlib provides only image utilities)
vcpkg install dlib[core] onnxruntime --triplet x64-windows

# Build (FaceLoginConsole.exe is written directly to installer/FaceLoginSetup/resources)
cmake -B build -S . -G "Visual Studio 17 2022" `
    -DCMAKE_TOOLCHAIN_FILE="$env:VCPKG_ROOT/scripts/buildsystems/vcpkg.cmake"
cmake --build build --config Release

Go Installer

cd installer/FaceLoginSetup
# Sync C++ binaries and models into resources/ first, then build the installer
.\build-installer.ps1

build-installer.ps1 first builds a lightweight standalone Uninstall.exe without embedded installation resources, copies it into resources/, and then builds the full FaceLoginSetup.exe. Direct Wails commands remain useful for local iteration, but releases should use this script.

Model Files

File	Purpose	Download
2d106det.onnx	106-point face landmark extraction	InsightFace
det_500m.onnx	SCRFD face detection	InsightFace
w600k_mbf.onnx	InsightFace face recognition	InsightFace
minifas_quantized.onnx	Silent anti-spoofing	facenox/face-antispoof-onnx
head_pose_mobilenetv2.onnx	MobileNetV2 head-pose estimation (Yaw/Pitch/Roll)	yakhyo/head-pose-estimation

---

Contributing

Issues and Pull Requests are welcome!

• Code style: C++20, /W4 warning level
• Make sure the build passes before submitting
• Discuss significant changes in an issue first

---

Open Source License

MIT License © 2026 美国伐木工&EthanZer0

---

Disclaimer

This software assists Windows sign-in through face recognition but does not replace your password. Face recognition is a convenience feature; the system always keeps password sign-in as a fallback. Do not rely on face recognition alone in environments with high security requirements.