GunduLabs/gaze

Facial authentication for Linux

What it solves

Gaze provides a secure, local facial authentication system for Linux users, replacing or supplementing traditional password entry for logging in, unlocking lock screens, and executing sudo commands.

How it works

It operates via a background daemon (gazed) that communicates over DBus. When an authentication request is triggered, the system captures a camera frame, detects and aligns the face using the SCRFD model, computes a face embedding via ArcFace, and verifies liveness using MiniFASNet-V2 to prevent spoofing. All processing and embeddings are stored locally on the device.

Who it’s for

Linux desktop users who want biometric authentication integrated into their system's PAM (Pluggable Authentication Modules) stack, specifically those using GNOME, KDE Plasma, or Hyprland.

Highlights

  • System-wide Integration: Integrates with PAM for login, sudo, and lock screens.
  • Anti-Spoofing: Includes local liveness detection and support for infrared (IR) cameras.
  • Privacy-Focused: All face recognition and data processing happens on-device.
  • Hardware Acceleration: Supports CPU inference and optional OpenVINO for Intel GPU/NPU acceleration.
  • Desktop Environment Support: Includes dedicated extensions for GNOME and KDE Plasma.

相關

  • 專案
  • 專案
  • 專案
  • 專案
  • 專案