Summary
I collaborate with Bellingcat's open-source team to work on parts of the ENF matching process. I've started collecting European ENF data. You can check out a full writeup of the approach in my article, but the gist is that I found a server to misuse and started collecting ENF data from it. These data will be published when I figure out how to, but for now I'm working on the analysis part — and learning a lot! There's not much to show for the analysis part, but the approach is to use K-means clustering, dynamic time warping, and other techniques to match ENF data to audio samples — which allows for a precise chronolocation of the audio, as well as an element of verification (if the ENF data doesn't match, the audio is likely fake).
Notes
This project is a work in progress — check it out on GitHub!
I'll update this page as I make progress.