eyMate

Linux face detection similar to Windows Hello and Howdy written in rust.

[!WARNING] This project is very much new. It should work, but expect some stuff not to work as expected. IR IS CURRENTLY NOT WORKING CORRECTLY (only normal cam works)

How to Install

EyMate provides following package formats (and build commands):

[!NOTE] Only the pacman is offically maintained by me (I use Arch btw.), other (rpm and dpkg) might not work.

• pacman: cargo aur
• rpm: cargo build && cargo generate-rpm
• dpkg: cargo deb

The packages will be in a subfolder from 'targets'.

Alternative download the packages from the github releases.

PAM configuration

[!NOTE] Without the pam entries eyMate won't work. PAM config files are for specific auth types. You will most likely change Following entries:

• sudo (sudo prompts)
• kde (KDE screenunlock)
• polkit-1 (GUI prompts)

[!WARNING] Adding eyMate login to any display manager (sddm, gdm, etc.) is NOT RECOMMENDED! It will cause bugs (like with kwallet) and will make you system very unsecure!

Add following to pam config at the top of the file:

auth        sufficient      pam_unix.so  try_first_pass likeauth nullok
auth        sufficient      libpam_eymate.so

This will cause you login prompt to try face detection whenever you type in a wrong password. Just hit enter when you want to use face detection.

Alternative you could do it the other way around:

auth        sufficient      libpam_eymate.so
auth        sufficient      pam_unix.so likeauth nullok

This will first eyMate to do the auth and if it fails run a normal password login.

[!NOTE] Be aware that the second method will instantly cause the face detection to run and might be a security issue (unintentional logins).

Manual Install

Build project with cargo build --release.

You will find the executable and pam lib in the 'target/release' folder. Copy the 'eymate' executable to '/usr/bin/' to install globally (optional, only needed for cli). Copy the 'libpam_eymate.so' libary to '/usr/lib/security/' (required for pam auth). Copy from the 'prebuild' folder the 'vggface2.pt' file to '/etc/eymate/' (required by cli and pam auth).

Dependencies

You need to have following installed:

• pytorch

• torchvision

• python-gobject

• opencv

• pam

• facenet (git clone into project root, only for building vgg2face.pt)

Build vgg2face model

Just follow these steps:

• install dependencies
• run build_model.py
• copy created model ('vgg2face.pt') to '/etc/eymate/'

How to use

Configure

Configure eyMate with the config file found in '/etc/eymate/config.toml'

[video]
mode = "IR"         # Mode IR (ir cam) or RGB (normal cam)
device_rgb = 0      # Normal cam number /dev/video0
device_ir = 2       # IR cam number /dev/video2

[detection]
min_similarity_rgb = 0.7    # Min Sinilarity for face detection (lower means easier detection but worse for security)
min_similarity_ir = 0.9     # Same as above, but for ir cam
min_brightness_rgb = 50.0   # Min brightness for face detection (to prevent bad frames/detection)
min_brightness_ir = 10.0    # Same as above, but for ir cam
retries = 10                # Amount of retries befor auth fails

Add User

Add user with following command:

eymate add <user>

Test User

Test user with following command:

eymate test <user>

User must be added first, otherwise test will fail.

FAQ

IR Cam not working / all black

You might want to install and enable linux-enable-ir-emitter, so your IR emitter turns automatically on.

Can I contribute

Of course feel free to open issues and pr's at any time. However note that you should open a issue for bigger features first, to prevent work duplication and misconceptions.

Disclaimer

[!WARN] Use this software at your own risk! It is most certainly bad for security, so don't expect your system to be safe with it. I am not responsible for any damage done or cause by this software!