Midnight Enigma dev page

[Curtis Vaughan (ACSS)]

Question about the ME dev page. There are Broken ciphers for May and June. But the links to the challenges and cookbooks lead to 404 errors. Maybe these are just examples and were never supposed to be used, but if there are such challenges I would like to see them.

[Jürgen]

Indeed, the Midnight Enigma page is still a placeholder at this time, meant to solicit feedback about the format of the page and the competitions. Oscar has already made it much more functional than I expected, but there is still some dummy content. 

Maybe Oscar can chime in on the planned next steps. And if anyone has feedback on the current draft, feel free to let us know!

[Curtis Vaughan (ACSS)]

Well, on the How-To page for ME, there is a section of recently submitted messages. Are these legit and should one take a crack at deciphering them? Or are they just dummy messages.

[Jürgen]

I'll let Oscar chime in, who has taken the lead on the Midnight Enigma page. But I believe the "Recently Transmitted Messages" at the bottom of the page may have been encrypted with an actual Enigma simulator. But even the "broken" ones, where most key settings are give, do not specify the rotor start positions. (Or may specify them, mislabeled as "Ring" settings, but then lack the actual ring setting?) So it's still not an easy task to decipher them. 

The historical Enigma messages found on the internet are mostly (all?) German plaintext for obvious reasons. I can understand that non-German users are typically not keen on that additional level of deciphering... so we better get Midnight Enigma off the ground soon!

For what it's worth, when I asked ChatGPT about Enigma practice messages in English (hoping for some web references), it generated some for me! The key settings looked complete and plausible right away, for an Enigma I, including the historically correct 10 plugboard connections. But decoding the messages initially resulted in gibberish. When I complained about that, it generated new messages which decoded correctly!  

I am not generally a fan of "look what AI told me" posts, but was quite impressed by this. When asked, ChatGPT explained that it had programmed an Enigma simulator in Python behind the scenes to generate the example messages. So, if you are into this kind of stuff, it's a neat way of playing with old and new technology at once.

Cheers,

Jürgen

[Jeff Spears]

ChatGPT could not (thankfully) decode my message:

<jef>

Can you decode the following  german enigma code:

AUG 14 0630Z = 80 = DEC RNF

XEHXI YIBJH VKQTO ZBXVL OEHQB DYSOB KKUTM IPPET MAVKL
NBIWF OZQYP KEBXF PWECR MEOTK HZCXN IDFFQ
ChatGPT said:

</jef>

<ChatGPT>
The message you're asking about seems to be related to a German Enigma code, a cipher that was used by Nazi Germany during World War II. The Enigma machine was a complex encryption device that used a series of rotors to scramble messages, and cracking its code was a monumental task that played a crucial role in the Allied victory.

Unfortunately, decrypting an Enigma message directly is not possible without the exact settings used on the machine (like rotor configurations, plugboard settings, and the daily key). However, I can help explain how this code might have been decrypted historically.

The Enigma machine worked by setting up a variety of parameters for encryption, and once those parameters (like the rotors and plugboard settings) were set, they would create a completely different cipher each day. The message you’ve posted ("AUG 14 0630Z = 80 = DEC RNF" seems to be indicating a specific date and possibly a key or message identifier) appears to be an Enigma-encrypted message with no obvious modern decryption tool available without the exact machine settings.

If you have any additional information, such as:

    The key or settings for the Enigma machine on the date in question (like rotor positions, plugboard connections, etc.)

    If this message corresponds to a known cipher, historical event, or if you have any related context, that could help narrow down a more accurate decryption.

Without those specifics, decoding Enigma messages is largely a historical challenge unless you're using computational methods or have detailed machine settings.

Let me know if you want to dive deeper into how the Enigma worked or if you have more information that might help!

</ChatGPT>

You frightened me there Jurgen! :-)

[Jürgen]

• ChatGPT could not (thankfully) decode my message:
  

• [...]

• 
  You frightened me there Jurgen! :-)

Yep, ChatGPT is not competent as a codebreaker yet -- phew! And it needed some help even when I tried giving it they key settings: 

It can't read the scanned 2744 key sheet properly: could OCR the letters and understand the column format, but lost track when trying to extract the settings for the 14th. And when I gave it the proper machine settings for the day, it still needed help with the crypto procedure: It assumed the old approach of encrypting the message key twice, and was not aware that the first group of 5 was not part of the ciphertext.

So we still have an edge over it even as Enigma operators -- reassuring!  ;-)

[Curtis Vaughan (ACSS)]

Inspired by Jurgen's adventure into using an LLM to create practice messages, I did the same using Gemini (since the full version is provided through my job). 

What a pain! I spent a lot of time trying to get it to work. Of course, it's possible I'm making an error, but whatever it came up with just resulted in gibberish. Despite constantly working with it to point this out, I was never able to get it to work. 

It is quite interesting though to watch it test itself by showing the conversion of a letter through each stage of the Enigma. I suppose I should see if there is a simulator that also provides such info, so that I could then point out to Gemini where it's going wrong. Anyhow, it's been interesting.

I also have ollama localized on my computer with the codellama module. As is typical for ollama and it is always very confident in itself and makes up stuff. It insisted that there were 2 Windows-based simulators that I should use, but all searches did not discover the apps it was suggesting. Also the parameters it would state I needed to use where preposterous. Oh well......

[Curtis Vaughan (ACSS)]

Well, I'm miffed. When I was working with the LLM on practice messages, I found the following website and was using it:

https://people.physik.hu-berlin.de/~palloks/js/enigma/enigma-u_v26.html

But now I've realized that even messages that work fine on the ET, just come out as gibberish on that website. 

[jue...@e-basteln.de]

Hi Curtis,

That should not be the case! I have found the Palloks simulator to be very reliable.

It does have a few gotchas in its UI though. Have you found the slightly hidden Ring settings, and noticed that it is not sufficient to enter the plugboard connections, but you must then "Activate" them?

I am pretty confident we can get the Enigma touch and the Palloks simulator to agree!

Best,

Jürgen 

---

Von: enigma...@googlegroups.com <enigma...@googlegroups.com> im Auftrag von Curtis Vaughan (ACSS) <cava...@avantguardsystems.com>
Gesendet: Montag, August 11, 2025 4:56:43 PM
An: Enigma touch <enigma...@googlegroups.com>
Betreff: [Enigma touch] Re: Midnight Enigma dev page

--
You received this message because you are subscribed to a topic in the Google Groups "Enigma touch" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/enigma-touch/ra3oMJlC7Io/unsubscribe.
To unsubscribe from this group and all its topics, send an email to enigma-touch...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/enigma-touch/3dce6dbc-1cab-47bc-aba8-d92ee82de30an%40googlegroups.com.