Engineering your own voynich

[nablator]

Yeah I know, it's not entirely optimal. But at that point I gave up

I made my own similar cipher to see if I could solve it knowing the method and the parameters for sure (space + 26 letters, two cuts per ciphertext word):

Code:

tiiaoean aoooan yiielioa fiao qaaaenuy tuiib taaoool tiiioa teal qeioa nuegf temouy bumean hib qiaiean yiinn buelben yepfitioa tiimean qeioa qeal qeoan tiiaoean taparioa teaii tuellen qeemiunh qeunh taaooq lieliea aiiyun temean qaouy fvameean tafitean hiq qial tet qiyonuy liaeiinn fviiean aminn fvt fvegn tuei qiel yepaoioa fiao lim liii yepfitioa tiimean qeinn buii daiiean qeb bufitar qiel hiaaeq fvai tiiemi tiiaoean aaiial taegyun aameioa gamean fvelq kib fvelq tufityun aaaenuy hipen tiiyon liiiq liaei buelal nuyopen liii lielben liaeial tiiaiipen aaooq aeial yiiinn buooean aaeeean fvel fiemino qoemipen fvai liel limal nuyo hioouy lieliea aminn fvioa tiiaaen liaaenuy afitean qeaaeioa teiial qeaeiean aminn fvoan taaooiea lielpen aaaef tiiaooan taiiinn liema buelben aemaf taemaoan teaii qiame tiiaoean aeial qeopen fifitioa qom bufitean aaiean daaeq fvai uaaeouy liemif teinn yiiemaoan tiiaoean teaiipen aaoal uaii daiiean qepen tuamiiea fviiq fvai lielb nuoben lieliea aminn bum lim liii buel qiminn fvfitunh liiiean aonuy taaiib qifitben fvai teaeiouy liemaq aameoan taal aelal tiioouy bueg nuaaenuy koal kaii qifit fieminuy temouy tuaeiq fvai nuaaenuy koal kaii buel tuelouy fvaaeben aaeioan hiaii fiyool kiaiiouy aemiouy aemiq kiaiiouy aameoan taei qiame fifityun qomoan kfital qoaoyun aemiouy aiiioa qaeman yiial aeiean buelioa qeamin fviipen aaooan bux

It's harder than I thought. A simple hill climb algorithm leads to sub-optimal outcomes (ambiguities or too many letters). Simulated annealing is either extremely slow or doesn't reach the optimum.

Does anyone have an idea for a better algorithm?

EDIT 1: One possible idea for improvement is to prioritize moving cuts in words that produce the highest number of ambiguities.

EDIT 2: Reading the comments for the first time here: You are not allowed to view links. Register or Login to view.

Mark Knowles on You are not allowed to view links. Register or Login to view. said:
I should probably say that if one has a triglyph
$&#
Then one shouldn’t also have the biglyph
&#
And similarly not
#

Actually this is not a necessary condition for a deterministic ciphering and deciphering. Having $&# and $& and $ at the same time is also possible.

* * *

Now I realize one of the possibly reasons for having "too much structure" in a ciphertext: spaces can be optionally added/removed in many places to make it even more difficult to crack.