I must admit I have never understood how to really use the ssh fingerprint visualizations. They look just like a random mess to me and I'd guess I'd never notice some smaller mismatch.
I did not know the drunken bishop story, but even knowing it now does not seem to help. As the author writes, the visualization of density does not seem to work very well.
If you're wondering why there's a "drunken bishop" reference: There's a little story in the intro to our paper to explain the OpenSSH fingerprint visualization algorithm. http://dirk-loss.de/sshvis/drunken_bishop.pdf
What language is that? One of the tags says llvm-ir. Well, I know what LLVM is and I guess IR is its intermediate representation between frontend and backend.
Would one program in that manually? Why? Because it can be done or some better reason?
I'd guess because of "check" from Cyclic Rendundancy Check. It's not an actual CRC, and the check is intended for human eyes rather than for a machine, but it's still a checking tool.
oops- I'd have figured there was some kind of hash process between the key and the image, so that a tiny change in the key would result in a completely different image, making the combination of key and image one step more difficult to closely duplicate, but this doesn't appear to be the case
I think for humans, something like https://robohash.org/ would be much better. An ASCII-style port for the console should be possible.