The differences he's looking at are digital breadcrumbs left behind by editing software. It's not so much an analysis of the photo as it is of the JPEG file.

The differences he's looking at are digital breadcrumbs left behind by editing software. It's not so much an analysis of the photo as it is of the JPEG file. JPEG is lossy, and repeatedly saving the same image changes the data in a predictable way, such that it's possible to identify regions of "newness" in a photo based on how their compression error differs from other parts of the same file.