New M4 batch - U-534 P1030680 |
Message boards : News : New M4 batch - U-534 P1030680
Previous · 1 · 2 · 3 · 4 · Next
Author | Message |
---|---|
matajan Send message Joined: 19 Apr 10 Posts: 6 Credit: 851,215 RAC: 0 |
In my last post I explained one of the reasons: "According to my tests, an IC of 0,044 IS A GOOD IC. The current algorithm despises this (LOW) IC and try to find the highest IC possible." Another reason is the search for the greatest score value using trigrams. For example, the triagram ein has the highest score value and the algorithm search the highest sum of triagrams score value so that the ideal solution for it, would be always eineinein ciphertext, ignoring any other solution. It is a complex and delicate problem, this text is good or just garbage? so I think we need a smarter algorithm. Best regards, matajan. |
TJM Project administrator Project developer Project scientist Send message Joined: 25 Aug 07 Posts: 843 Credit: 267,994,998 RAC: 0 |
I think the algorithm is pretty good considering the fact that it's used on a very short text. It also gets stronger as the message length increases. With strong dictionary it easily breaks 72 letter ciphertexts. The problem is the fact that we don't know if a dictionary is good until the message is broken. Also, I doubt you actually could get a text like 'einseinseins...' or something like that. For a random output, if the score increases on one part, the stecker gets messed up badly and the score of the rest of the message drops by a lot. Surely there are 'local maximums' where the score goes up far beyond the average on a random decrypt (as can be seen on main page - top results of 1,9M where the average is just 1,6), however if the trigram dict fits the text, the plaintext score will be even higher. M4 Project homepage M4 Project wiki |
matajan Send message Joined: 19 Apr 10 Posts: 6 Credit: 851,215 RAC: 0 |
Obtain a text as: "eineinein..." it is practically impossible. I meant a possible text formed with trigrams with a high value as garbage like "ein...von...die...ich..." I just take a look at dict files, why exists a lot of entries with zero value? This only increases the processing time and is totally useless. Best regards, matajan. |
TJM Project administrator Project developer Project scientist Send message Joined: 25 Aug 07 Posts: 843 Credit: 267,994,998 RAC: 0 |
I think you are wrong. The bigram/trigram arrays have constant size of [26][26] and [26][26][26], so there is no performance loss other than reading more data from file. M4 Project homepage M4 Project wiki |
matajan Send message Joined: 19 Apr 10 Posts: 6 Credit: 851,215 RAC: 0 |
Sorry, you are right. I am seeing that are obligatory all combinations of 2 and 3 characters because the function implementation used all the combinations as index although its value is zero. I think that optimizing these sum functions could increase the processing speed. What do you think about this? |
TJM Project administrator Project developer Project scientist Send message Joined: 25 Aug 07 Posts: 843 Credit: 267,994,998 RAC: 0 |
Well, recently I've been testing highly optimized app for modern CPUs (beats any earlier versions), perhaps it will be released as default via plan_class. The project status for today is: wu_filename - dict - avg/min walks: vroln72_1 - naval - 37/33 vroln72_3 - u534 - 60/54 fnygm76_1 - naval - 21/20 fnygm76_2 - u534 - 20/19 The processing rate (on average) is near 4k workunits/hour, which results in over 2 full walks over keyspace/day. I'll post top results later today. http://www.enigmaathome.net/static/3356xwn1mw/vroln721.txt http://www.enigmaathome.net/static/3357x13h75/vroln723.txt http://www.enigmaathome.net/static/3354x91s9m/fnygm761.txt http://www.enigmaathome.net/static/3355xwtqe6/fnygm762.txt M4 Project homepage M4 Project wiki |
TJM Project administrator Project developer Project scientist Send message Joined: 25 Aug 07 Posts: 843 Credit: 267,994,998 RAC: 0 |
I think that the 76 letters long FNYG MXHU is broken. m4_fnygm76_2: score: 1902866 settings: C:FB:ZPEU:G528:CPDGEJFIKTLZMSNOQURW leitungvvvuuustuetzpktxwwwhavenxxfffttteunszwozwovierhuermitvrrhhhvvvgeloest M4 Project homepage M4 Project wiki |
TBREAKER Send message Joined: 26 Sep 11 Posts: 29 Credit: 0 RAC: 0 |
CONGRATULATIONS!!! You made it. A new world record, I guess. 76 letters! AN LEITUNG VON U BOOT STUETZPUNKT WILHELMSHAVEN: FUNKTELEGRAMM EINS ZWO ZWO VIER HIER MIT RHV GELOEST. Translation is: To administration from U Boot Command (Headquater) Wilhelmshaven. Radio message 1224 solved with RHV (Hand Cipher System)... GREAT!!! -=> Breaking German Navy Ciphers - The U534 Enigma messages <=- |
TJM Project administrator Project developer Project scientist Send message Joined: 25 Aug 07 Posts: 843 Credit: 267,994,998 RAC: 0 |
Thanks ! It was the 48th walk over the full keyspace, so lots of CPU time went into this. Thanks for everyone who participated! M4 Project homepage M4 Project wiki |
TBREAKER Send message Joined: 26 Sep 11 Posts: 29 Credit: 0 RAC: 0 |
Ok it is message P1030713. I thought it was 680 ;-) -=> Breaking German Navy Ciphers - The U534 Enigma messages <=- |
TJM Project administrator Project developer Project scientist Send message Joined: 25 Aug 07 Posts: 843 Credit: 267,994,998 RAC: 0 |
The message I posted above is actually the P1030713. P1030680 and P1030713 have been running in parallel due to limitations of the server software - there is no way to abort all work that was already sent to the clients. Having a 110-120k work units in progress at any time, which is close to 4 full keyspace walks, hurts a bit when the message is broken and suddenly they're not needed anymore. Running two text instead of one reduces the 'wasted' CPU time by 50%. M4 Project homepage M4 Project wiki |
krzyszp Send message Joined: 2 Sep 07 Posts: 1 Credit: 1,952,259 RAC: 0 |
|
TJM Project administrator Project developer Project scientist Send message Joined: 25 Aug 07 Posts: 843 Credit: 267,994,998 RAC: 0 |
A status update on the P1030680: I paused the m4_vroln72_1 batch, which uses old "naval" dictionary, because I have a feeling it might not work too well. It fails on ~40% of U534 messages. Very few are completely unbreakable, but there are plenty of random 'decrypts' scoring higher than the plaintext. Currently the server sends out only m4_vroln72_3 based on 'u534' dictionary set. I think I'll add a second batch which will use updated dictionary, I'm not sure yet if it will replace the current m4_vroln72_3 or if I'll let it run in parallel. The recent break brought more participants and currently the average processing rate is a bit over 2 full walks/day, which is very impressive. M4 Project homepage M4 Project wiki |
rilian Send message Joined: 30 Jan 09 Posts: 5 Credit: 224,941 RAC: 0 |
TJM, is it right that just only one message is remaining unbroken on Enigma@Home ? http://enigma.hoerenberg.com/index.php?cat=Unbroken&page=P1030713 |
TJM Project administrator Project developer Project scientist Send message Joined: 25 Aug 07 Posts: 843 Credit: 267,994,998 RAC: 0 |
|
TJM Project administrator Project developer Project scientist Send message Joined: 25 Aug 07 Posts: 843 Credit: 267,994,998 RAC: 0 |
The new batch m4_vroln72_4 is assigned to r1u534 dictionary set, which is a revised version of the previous u534 set. Not much has changed, but on a few tests I ran, the new revision seems to produce better results. M4 Project homepage M4 Project wiki |
CoolerMaster Send message Joined: 30 May 13 Posts: 1 Credit: 30,525 RAC: 0 |
I signed up because of the main text that described how I could help break the remaining Enigma messages. But after several weeks I've discovered that they have all been broken and now the project is taking on many other tasks not related to the original goal. I didn't sign up for that and am very disappointed that the main page no longer reflects the current status of the project, that's misleading. Therefore I'm discontinuing my help. |
dedo Send message Joined: 14 Sep 13 Posts: 2 Credit: 210,593 RAC: 0 |
I signed up because of the main text that described how I could help break the remaining Enigma messages. But after several weeks I've discovered that they have all been broken and now the project is taking on many other tasks not related to the original goal. I didn't sign up for that and am very disappointed that the main page no longer reflects the current status of the project, that's misleading. Therefore I'm discontinuing my help. I have signed up recently too, very nice project, but I'd like to know which messages have been broken and what we are working on right now. Is there a page somewhere that shows this information? |
dedo Send message Joined: 14 Sep 13 Posts: 2 Credit: 210,593 RAC: 0 |
It looks like we are working on this right now. http://enigma.hoerenberg.com/index.php?cat=Unbroken&page=P1030680 |
Thomas Hansard Send message Joined: 23 Nov 13 Posts: 6 Credit: 595,836 RAC: 0 |
Is the current subject of this project P1030680? |
Message boards :
News :
New M4 batch - U-534 P1030680