IRA Cipher Traffic
Around 1,300 cryptograms from the papers of an IRA chief of staff, sealed in an archive for seventy-five years, and broken almost in their entirety by one cryptanalyst with a hill-climbing program.
Why This Matters
This exhibit is about a body of real ciphertext and the man who read it. In the 2000s the historian Tom Mahon found roughly three hundred documents containing encrypted messages in a Dublin archive — the papers of Moss Twomey, IRA chief of staff from 1926 to 1936. Between them they held about 1,300 individual cryptograms, mostly dispatches between headquarters and organisers in Britain and the United States, dating from 1926 and 1927 with a scattering from 1925 and 1928. Mahon brought in the American cryptanalyst Jim Gillogly, and together they published Decoding the IRA in 2008. All but one of the cryptograms are now readable.
The traffic belongs to the years after the Irish Civil War, when the organisation was rebuilding and operating across the Atlantic. The enciphered dispatches are easy to spot in the file even before they are read: blocks of five letters carrying no visible meaning.
What broke one message was not cryptanalysis but filing practice. Alongside the ciphertext, headquarters kept a précis — a plain-language summary — of what had been sent. In one case Twomey wrote to Connie Neenan in America asking him to "send Cooney [a] copy of the addresses he had", and then summarised it in the file as "send address to Smith". The plaintext and the ciphertext sat in the same archive.
The system was columnar transposition. The message is written into a grid of fixed line length; the columns are then read out in an order set by a keyword. No letter is ever replaced — every letter of the plaintext survives into the ciphertext, only its position changes. That is why frequency analysis alone is useless against it, and why the letter counts of a transposed message still look like ordinary English.
Operators also padded. In the message Gillogly works through as an example, the encipherer had inserted two columns of meaningless vowels to disguise the shape of the grid.
The demo above uses the museum's columnar engine. It is the right mechanism, but it is a clean-room model: it does not reproduce the padding conventions, the five-letter grouping, or the handwriting of the originals.
Gillogly wrote hill-climbing software aimed at columnar transposition. For each message he tried line lengths in the range of about eight to fifteen characters, ran the search around a hundred times from different random starting keys, and watched for the point at which recognisable English words began to fall out of a candidate solution. The approach is the same family this museum covers under hill climbing, and it is what makes a hand cipher that was entirely adequate in 1926 transparent today.
A single 59-letter cryptogram in the collection has never been read:
GTHOO RCSNM EOTDE TAEDI NRAHE EBFNS INSGD AILLA YTTSE AOITDE
One suggestion that has been raised about it is that the underlying language may be Irish rather than English — which would defeat a search scored on English text. That remains a suggestion, not a finding.
| Source | Moss Twomey papers, UCD Archives, Dublin |
| Era | Mostly 1926–1927; a few 1925 and 1928 |
| Family | Columnar transposition |
| Corpus | ~300 documents, ~1,300 cryptograms |
| Discovered by | Tom Mahon (historian) |
| Broken by | Jim Gillogly, hill climbing over column orders |
| Published | Decoding the IRA (Mercier Press, 2008) |
| Still unsolved | One 59-letter message |
| Modern Lesson | Transposition hides position, not letters — and a filed plaintext summary undoes both |