The Enigma messages AWTZK and ZNLZT from 11 July 1941 are among the few remaining hard Enigma messages that have resisted cryptanalytic attacks since they were made publicly available around 2005.
Since Carter Leffen broke the MVUEH message with help from OpenAI GPT–6 Astra on 15 September 2026, interest has grown in breaking the remaining Enigma and Truppenschlüssel messages on my webpage, Crypto Cellar Research. On 24 September 2026, Sean Harrington contacted me and said he was trying to attack the two Enigma messages AWTZK and ZNLZT. He asked whether I would be willing to share the plaintext for the Truppenschlüssel (TS) messages from the same date, 11 July 1941. Mr Harrington was looking for cribs to use in the attack, and we felt that if the TS plaintext could help, we would gladly provide those. Olaf Ostwald dug into his files and provided the TS plaintexts for 10 and 11 July 1941.
Big was my surprise when I received an e-mail from John Woodhouse late in the evening on 26 September 2026, announcing that he had broken the two messages AWTZK and ZNLZT. A quick look at his result convinced me he had indeed broken the two messages, but I waited until the next day to make the break public and to allow myself to check the decipherment. Mr Woodhouse’s break is very interesting, and it shows why these two messages have resisted breaking for so long. The break shows that the ciphertext of the 49-letter AWTZK message has 9 garbles, while the 69-letter long ZNLZT has 11 garbles. That our hill-climbing attack failed on these rather short messages with so many garbles is not so surprising. You can read his detailed account of the break here Break of Enigma messages AWTZK and ZNLZT.
I was sorry to inform Sean Harrington that somebody else beat him in the race to break AWTZK and ZNLZT. Luckily, Mr Harrington took the bad news well, even if he said: “It’s a little painful, since we were working on exactly those two messages.” After I had posted the daily key for 11 July 1941, Sean Harrington had another look at his logs and discovered that he had been very, very close. Here is his report: “The GPU search had the right wheel order (I IV III) and at its best 4 of the 10 plugboard pairs, but the hill climb never reached the rest. My check would have accepted the true key if the climb had found it, but at 49 and 69 letters the climb didn’t converge.” This is not a unique case; already in 2005, when we did the initial breaking, we often found that a run stopped just before it found the key.
Updated: 27 September 2026 at 19:13 UTC
| Back to Crypto Cellar Research | Licensed under CC BY-NC-SA 4.0 Frode Weierud, Crypto Cellar Research © 2026 |
Breaking German Wehrmacht Ciphers |