Hello,
We have not published the dataset for the CoralNet 1.0 feature extractor, a big reason being that the dataset did in fact include private sources. There also wasn't any particular consideration to pick more public sources than private sources. In fact, with the ratio of public vs. private sources at the time, only 10% or so of the chosen sources were public.
I was digging around for the set of 304 source IDs mentioned in the paper, but I was only able to find a list of 284 which might have been from an old iteration. Still, to at least give you a general idea, here are the 27 sources from that list of 284 which are currently public:
23, 70, 109, 155, 258, 295, 307, 350, 372, 373, 376, 450, 466, 503, 554, 616, 620, 793, 800, 841, 842, 843, 1073, 1076, 1288, 1388, 1579