(note the inspiration was a nefarious regex scanner for finding bitcoin hashes. I have no intention of building such a thing, but the idea of a random detector regex intrigued me. Is it possible?)
My idea was more like calculate the statistics of character n-grams in english (all of which exist in random noise too), but count the most unlikely occurrences until you hit some probabilistic threshold that indicates some well thought out decision boundary.