TLDR; I had some fun playing around with combining columnar transposition with substitution, which turns out to be surprisingly strong. Almost certainly not "properly strong", but not trivial to break with computers either. I was really surprised by this and would love to heart your ideas how this might be attacked more properly. I'm a physicist with background in stats & ML and an interest in cryptography and I got nerdsniped pretty hard a while ago by this quote on wikipedia For example, a simple substitution cipher combined with a columnar transposition avoids the weakness of both. I've always found doulbe columnar transposition pretty interesting because Lasry's 2014 paper came out while I was learning about crypto in school. As far as I can tell, outside of special cases, the methods from that paper and their 2016 follow up paper are still reasonably representative of the state of the art. Though scoring functions and key generation have improved.…