*Given*: columnar transposition ciphertext *y*;

*Find*: the transposition width *N* and transposition *τ*.

Our plan is to test *N* as a possible width by computing a *Marko score* for the adjacency of columns in the ciphertext, assuming each of the *N*! transpositions of width *N* are equally likely to have been used.

Testing a width *N* is formulated as a hypotheses *testing* problem; for each pair (*i*, *j*) with *i* ≠ *j*, decide which of the two hypotheses is the most likely to be true.

, | jth column is read from X immediately after theith column is read from X. |

, | jth column is not from X immediately after theith column is read from X. |

When ADJ(*i*, *j*) is true, the *i*th and *j*th columns must be columns (*k*, *k* + 1) in *X* for *some k* with 0 ≤ *k* < *n*−1. As the *N*! transpositions *τ* have been chosen with equal probability, the *a priori*^{4} probabilities of the hypotheses ADJ(*i*, *j*) and are

and

The ratio of these probabilities is the *a priori odds* of ADJ(*i*, *j*) over

The term ODDS has the same interpretation ...

Start Free Trial

No credit card required