Generating key from your voice and face

I just posted an experimental repo exploring how we could use a persons face, voice print as well as a spoken secret phrase to deterministically generate a private key.

In this there, is a “notebook” of sample code which takes 2 pictures of a face, voice samples, and generates the same key twice from different source material with the phrase “my voice is my passport” (in reality - you want that phrase to be secret!).

You can explore the notebook here: experimental-face-voice-key/you_are_the_key.ipynb at main · TBD54566975/experimental-face-voice-key · GitHub

or play with it on google colab here: Google Colab if you are familiar with notebooks.

The aim here is that anything that we can do to make self-custody of private key material easier is good for SSI, and what could be easier than just being yourself in a way that it is near-impossible for anyone else to impersonate. This could be for key recovery if not for the key material itself.

Unfortunately, I am not an expert in any of these diverse domains, so any collaboration on voice prints, or facial biometrics and cryptography appreciated (this may not yield sufficiently secure results, but I am curios to see if is of interest to anyone).

Super cool! Damn, Python library support for these is powerful; look how much you got done in those lines of code.

This could be for key recovery if not for the key material itself…but I am curios to see if is of interest to anyone).

Definitely is. Key management is the riskiest and most inaccessible barrier keeping folks from using self-custody. I’d love to discover how voice or facial recognition is unique (to a person) and reliable (stable to use as recovery over time).

Brainstorming other identity types derived from you: LIDAR info, fingerprint…

Awesome stuff. For those with the expertise Mic’s looking to get involved: this is important and I’m here to support.

S,
ALR

one way for folks to contribute would be to have multiple face photos and voice recordings (the same phrase spoken multiple times, and some unrelated phrases to ensure we can differentiate) from people. With permission to use and publish them.