How the Columnar transposition cipher works

The idea in one sentence

Columnar transposition is a Transposition: every letter keeps its identity and only changes place. The keyword decides how the places are shuffled — write the message row by row into a grid, then read the columns out in the order the keyword spells.

The recipe

1. Choose a keyword, say SUN. Its letters, sorted alphabetically (N, S, U), fix the Column order: the column under N is read first, then S, then U.

2. Write the plaintext row by row under the keyword, one letter per cell: ATT / ACK / ATD / AWN.

3. Read the columns in keyword order, top to bottom: the N column gives TKDN, the S column AAAA, the U column TCTW.

4. The ciphertext is the columns concatenated: TKDNAAAATCTW. To decrypt, rebuild the grid the same way and read it row by row.

A worked example

Each row follows one plaintext letter: which column of the grid it was written into, and which position of the ciphertext it ends up at. The letters never change — A is still A — but their order is scrambled beyond recognition without the keyword.

Padding and the incomplete last row

If the message length does not fill the last row, the sender has two options. Padding completes the grid with a dummy letter (the tool uses X), which makes the rectangle regular — every column has the same length, and the ciphertext length betrays the possible column counts. Leaving the last row incomplete keeps the grid irregular: slightly harder to analyse, and the classic hand practice. The tool offers both.

Double transposition

Encrypting the result a second time with another keyword — the classic double transposition — shatters the simple column structure. The single-columnar weakness you will exploit below mostly disappears; that is why serious users (from WWI armies to the SOE) always applied the transposition twice.

Columnar transposition: guides

Guide 1 — your first transposition

1. Open the tool and type the keyword SUN. The grid shows the Column order as numbered badges: 1 under the column the keyword reads first.

2. Type MEET ME AT NOON and compare with the grid: the ciphertext is exactly the same letters, read down the columns in badge order.

3. Switch to decrypt mode and paste the cryptogram back — it returns unchanged. Transposition undoes itself when you know the reading order.

Guide 2 — padded or irregular

1. Encrypt a message whose length does not divide evenly by the keyword length — say 13 letters under SUN (3 columns).

2. Encrypt once with "Leave the last row incomplete" on and once off. With padding, the ciphertext length is a multiple of 3; without, it is 13 — the grid shape itself leaks or hides information.

3. Count how many candidate widths divide the padded length: each divisor is a hypothesis a codebreaker must try. Irregular grids cut that list down.

Guide 3 — double transposition

1. Tick "Double transposition" and add a second keyword, say WOLF.

2. Encrypt the same message with and without the second keyword and compare the two cryptograms: the double one has no readable column fragments left — the very fragments the single cipher leaks.

3. This is why armies doubled the transposition, and why the guided cracker in this app sticks to the single version.

Breaking the Columnar transposition

A transposition hides the order of the letters, not the letters themselves — so the frequencies that kill substitution ciphers are flat here, and a different weakness takes over: structure. The wizard on the Cryptoanalysis tab walks the attack; this page explains what each step does and why it works.

The flaw: the grid has a shape

The ciphertext is whole columns of a grid concatenated. If you guess the column count right, the ciphertext falls apart into exactly those columns — real, unbroken vertical slices of the plaintext grid. Guess wrong and you slice through columns at arbitrary points, mixing unrelated letters. That is the lever the whole attack works from.

The wizard, step by step

1. Column count. For a padded (complete) grid, the ciphertext length must divide evenly by the width — only the divisors are candidates, ranked by how English-like the best arrangement of their columns looks. The score is a hint, not an answer: on short texts a wrong width can outrank the right one, and the real test is step 2.

2. Order the columns. The ciphertext stands upright as columns; Anagramming means dragging them side by side until rows read like English. Each neighbouring pair is scored by the digraphs their rows form together — TH, HE, IN light up, QX does not. Watch the badges, move the promising partners together, and trust your eye for the language over the numbers: the tool never reorders a column by itself.

3. Done. When the rows read, the plaintext is recovered — padding letters may trail at the very end. "Use this result" carries the column order to the Tool tab, which decrypts the whole message.

Why it works: digraphs remember order

English letters are not independent: certain pairs are common, most are rare. Two true neighbouring columns zip their rows into real English digraphs; two strangers zip into noise. That single statistic is enough to rebuild the key one decision at a time — the same principle the ADFGVX cracker applies to its second stage, where the columns are the same object.

What double transposition changes

The Column order of a single transposition is recoverable column pair by column pair. Apply the cipher twice and no pair of ciphertext chunks is ever a true pair of plaintext neighbours — every simple statistic goes flat. Breaking double transposition by hand means writing out probable words on grid paper and sliding shapes around; glorious manual labour, and outside this tool's scope.