Welcome to biopass Discussions!

[phucvinh57]

👋 Welcome!

We’re using Discussions as a place to connect with other members of our community. We hope that you:

• Ask questions you’re wondering about.
• Share ideas.
• Engage with other community members.
• Welcome others and are open-minded. Remember that this is a community we
  build together 💪.

To get started, comment below with an introduction of yourself and tell us about what you do with this community.

[aszabo70]

Hi, I am testing it, on Debian, Fedora and Arch linux, so far everything works, but I would have an ask: could you enhance it with PIN authentication like in windows?

[phucvinh57]

PIN authentication is supported by hardware modules, like TPM or Secure Envlave. Theoretically, we can do it. However, there is only one form input to enter user password on most linux distros, so PIN method may conflict with the password one.

I will research it later, but #9 is a higher priority.

[aszabo70]

Thank you your answer. You know, I am using howdy and for PIN authentication libpam-pwdfile for years without any problems. Though the install of Howdy is not simple neither on Debian nor on any other linux distribution. About libpam-pwdfile you can find information for example here: https://randomblog.hu/setting-up-pin-unlock-for-ubuntu-linux/.
Thank you for your so far work.

[kunningKing11]

@phucvinh57 Hi, I've been using howdy for a while now until I got frustrated with the speed and the lack of active development. In comparison, this feels instant. I have only just started learning C++, but if there's anything you think I could do I would be glad to help.

Also, since this project moves so fast, it might be worth it creating a PPA for Ubuntu so that all updates to biopass happen automatically. I'm not sure if this helps, but ChatGPT told me this:

To create a Personal Package Archive (PPA) on Ubuntu, you must first register for a Launchpad account, generate and upload a GPG key, and create the PPA via the Launchpad interface. The process involves preparing a properly structured Debian source package and uploading it using the dput tool, as Launchpad builds binary packages from source rather than accepting .deb files directly.

Prerequisites and Account Setup

Launchpad Account: Register at https://launchpad.net and note your username, which forms the first part of the PPA URL (e.g., ~yourusername).
GPG Key Generation: Create a key pair using gpg --gen-key (recommended size 4096, expiry 2 years) and ensure the email matches your Launchpad account.
Key Verification: Upload your public key to Ubuntu's keyserver (gpg --keyserver keyserver.ubuntu.com --send-keys) and verify it via the confirmation email sent by Launchpad after adding it to your profile.

Creating the PPA and Preparing Packages

Activate PPA: Visit your Launchpad profile, click "OpenPGP keys" to link your key, then go to "Activate a PPA" to create a new archive with a unique, lowercase name.
Package Structure: Prepare a Debian source package with a debian/changelog file specifying the target Ubuntu release (e.g., noble for 24.04 or jammy for 22.04).
Build Source Package: Use dpkg-buildpackage -S -sa -k$KEY_ID to generate the .dsc and .changes files; ensure the .changes file is signed with the registered GPG key.

Uploading and Installation

Configure Dput: Set up ~/.dput.cf to point to ppa.launchpad.net with your specific incoming path (e.g., incoming = ~yourusername/ubuntu/ppa-name).
Upload: Execute dput launchpad ~/build/package_version_source.changes to submit the package; monitor the build status on the Launchpad page, which typically takes 10–30 minutes.
User Installation: Once built, users add the repository with sudo add-apt-repository ppa:yourusername/ppa-name, update their system, and install the package via sudo apt install package-name.

[phucvinh57]

No. In Windows Hello, it does work though but when I do ffplay /dev/video2 it looks quite dark which confirms my suspicions. ChatGPT told me that apparently it might be due to vendor-specific software that Linux might not have integrated.

Yep, the camera is vendored to be use in Windows only. That is bad. I have documented how to fix it in https://github.com/TickLabVN/biopass/blob/main/docs/IR%20camera.md.

Also for the liveness check in the biopass app I think it is a bit unclear what the difference between the AI detection and the IR detection is.

Anti-spoofing is a method that uses well-printed photos, recorded videos or 3D masks to fool the face authentication. To prevent it, we provide two ways:

1. Use an AI model to detect cheats. However, this method relies on RGB (colored) images taken from user RGB cameras. We are not 100% sure about its accuracy, in fact, it produces a lot of false positives and depends heavily on the light condition.
2. Use an IR camera to make sure a real person is sitting in front of their computer. An IR camera device has two components, a camera it self and its IR emitter (a small red LED). When using it, the emitter emits infrared light, which hits your face, then the camera captures the reflection of the light. If you show a photo of your face in front of an IR camera, you will get a gray rectangle, not your real face.

I will improve the documentation, thanks for pointing it out.

[kunningKing11]

Thanks for the information! I was now able to fix it with the link you provided. In terms of the documentation, I thought the most important thing to improve would be to show a link to the IR camera docs in the README.md or the PAM setup file. If you want to, just tell me which file to edit and I can make a PR for it.

[phucvinh57]

We have already included PAM link and IR camera link in Readme, but I think they are not very visible to users. You may edit them (includes the content of PAM/IR camera guides) if it is necessary.

[kunningKing11]

Done in #64

[kunningKing11]

Thanks for your recommendation, I will do it.

Any updates on this? It would make updating automatic every time someone runs sudo apt upgrade, so that more people are on the latest release.

[kunningKing11]

@phucvinh57 As mentioned earlier, since I am new to C++, forgive me if what I say does not make much sense but would it make sense to bump the C++ version from C++17 to something newer, like C++20 or C++23?
I did some research and it seems like C++17 is becoming "obsolete" and C++23 has newer features such as type safety which would make the code easier to read.

[kunningKing11]

Done in #64

[phucvinh57]

We used OpenCV as a dependency before. I do not remember exactly which version, but at that time, I got compile error with C++23. The OpenCV old version conflicts with C++23, so I just use C++17 for compatibility.

After that, I remove OpenCV because it is so heavy, which requires 1GB for installation. Now I think we can update to newer C++ version, but it is not our priority now.

Instead, I am looking a better way to implement camera-related functions. We eliminated OpenCV and switched to openpnp-capture, but openpnp-capture does not support IR camera. Currently, the IR camera function is being handled by raw V4L2 code, which is so complex, and I gave it up to Claude...