For pairs of files, I took their filename length, calculated the Levenshtein distance between them, their size in bytes and their duration in Ticks.

I used the ML.NET AutoML API to train a binary classifier.