News for hackers. Stuff that matters.
Re:Rare and hardest to crack Enigma code machine sells for $... by Animats on Monday July 20, 2020 @10:38PM (#23903873)
Four rotors. Now that would tire out your fingers.

The limitation on the number of rotors is key pressure. A 3-rotor Enigma has a huge key travel and you have to push hard. (I've had the opportunity.) The friction between the contacts is high. That's why these machines were not built with enough rotors to be really secure.

The next step up was the Lorenz 10-rotor machine, but that had to be motorized. But the crypto design of that was botched, and it's easy to crack. A real 10-rotor alphabetical Enigma-like machine without a "reflector" would have been secure against any attacks of the period.

Re:Rare and hardest to crack Enigma code machine sells for $... by tialaramex on Tuesday July 21, 2020 @04:16AM (#23905188)
> But the crypto design of that was botched, and it's easy to crack.

This seems like it depends upon a perspective unavailable at the time.

Until June 1943 (by which point Germany has in fact lost the war though it will fight on for a long while, the Eastern Front campaign is going terribly and the United States is increasingly diverting resources to a "Europe First" strategy which Germany can't conceivably overpower) the only breaks of Lorenz at Bletchley were based on what they called depth. Multiple ciphertexts using the same key, a fatal flaw for a cryptosystem even today.

For example a German radio operator begins encoding and transmitting an important message, and then is interrupted by some matter, after a few minutes they begin again but use the same key because it hasn't been impressed upon them that they absolutely must use a fresh key each time. If they make even a trivial typographical error the two ciphertexts will be different and that's enough to recover the messages.

Only in July 1943 do Bletchley begin actually decrypting any messages for which there is no depth. They do this because they have a (very primitive) electronic codebreaking machine, something that has never before existed. This machine is unsatisfactory in many ways but it's ludicrously faster than manual codebreaking.

So it seems unfair to critique Lorenz based on the fact that a suitably powerful electronic machine can break it. No such machines had ever existed before. Plus it took several years to even conceive of how to do the attack before the machine could be built, because of course programmable computers didn't exist yet.

Re:Rare and hardest to crack Enigma code machine sells for $... by taejo on Tuesday July 21, 2020 @04:46AM (#23905311)
> Multiple ciphertexts using the same key, a fatal flaw for a cryptosystem even today.

To avoid confusing readers, one should mention that modern ciphers separate the key, which can be reused but must be kept secret, from the initialization vector (IV), which must not be reused but can be transmitted in the clear.

These correspond roughly to the Enigma's daily settings and per-message rotor start positions, but AIUI the procedure for selecting and transmitting the latter changed over the course of the war, and at times insecure even if done according to the procedure.

Re:Rare and hardest to crack Enigma code machine sells for $... by stan_rogers on Tuesday July 21, 2020 @07:10AM (#23905927)
It was actually far worse for the Lorenz (Tunny) traffic. The wheel patterns - the equivalent of the key in, say, AES - were changed quite infrequently early on. The five χ-wheels (in Bletchley parlance) were only changed monthly, while the five ψ-wheels were changed quarterly. While the two μ-wheel patterns were changed daily, there was little consideration given to what those patterns actually were. The χ-wheels generated the first of two sets of bit patterns to be XORed with the plain text; the ψ-wheels generated the second set. That would have been fine, except that the two μ-wheels together determined whether or not the ψ-wheels advanced - μ1 advanced on every keystroke along with the χ-wheels, μ2 would advance only if μ1 was a 1, and the ψ-wheels advanced only if μ2 was a 1 - and bad patterns on the ψ-wheels would mean that the ψ-wheels would be essentially frozen for extended runs of the message. What was meant to introduce a sort of dither to make it harder to determine how the key was being generated became the Achilles heel that highlighted the mechanism. (Not to take anything away from Bill Tutte and the rest of the crew at all. There was still a lot of cleverness and tedious attention to detail involved.) With a month of life on the χ patterns and three months on ψ, depths and near-depths on the message settings (the nonce or IV) were absolute killers. Later on, all of the wheel patterns were changed daily, and the μ-wheel patterns were verified for short stall lengths, but by then the machine's mechanism was well understood, pattern-derivation methods had been devised, and Colossus had largely eliminated the χ component from the puzzle.
Re:Rare and hardest to crack Enigma code machine sells for $... by stan_rogers on Tuesday July 21, 2020 @10:11AM (#23906907)
The phrase "and bad patterns on the ψ-wheels" should have read "and bad patterns on the μ-wheels". Sorry if there was any confusion; I caught that too late for editing.
Re:Rare and hardest to crack Enigma code machine sells for $... by tialaramex on Tuesday July 21, 2020 @01:30PM (#23908849)
We're discussing (in this sub-thread) Lorenz rather than Enigma. But yes, the per-message rotor positions are what would be repeated resulting in "depth".
Re:Rare and hardest to crack Enigma code machine sells for $... by taejo on Wednesday July 22, 2020 @08:56AM (#23916396)
Yeah, I got side-tracked in the middle of writing my comment, and it didn't seem worth fixing since I only noticed after the edit window closed and my main point holds either way.
Re:Rare and hardest to crack Enigma code machine sells for $... by KMag on Tuesday July 21, 2020 @12:31AM (#23904330)
Instead of a reflector, they could have used 27 contacts, and either had a 27th lightbulb to indicate the output was the same as the input, or made each key a dual-throw dual-post switch so that the 27th output would get wired to light up the switch corresponding to the input key. That would remove the weakness that no letter is ever enciphered as itself.

Edit: never mind: you'd actually need 28 contacts, since the reflector can only reflect an even number of contacts. With an even number of unused contacts on the input side of the rotor stack, you could connect them together in pairs to get extra passes through the rotor stacks. If they added / (for use as a number shift symbol, so 1123 would be /AABC/ ) to their alphabet, then they could use 28 contacts and have something like described above. However, without modifying the alphabet space to have an odd number of symbols, all other workarounds I can think of require an odd number of used inputs at the one end of the rotor stack and and even number of contacts at the reflector end.

The Germans used Q or Z as their number shift symbol when enciphering digits for Enigma traffic, correct?

Re:Rare and hardest to crack Enigma code machine sells for $... by taejo on Tuesday July 21, 2020 @04:52AM (#23905341)
> The Germans used Q or Z as their number shift symbol when enciphering digits for Enigma traffic, correct?

In the few random decrypts I looked at (from U-Boat traffic), e.g., https://enigma.hoerenberg.com/index.php?cat=The%20U534%20mes... numbers were spelled out.