enigma-cuda

about

enigma-cuda is a command-line tool for ciphertext-only cryptanalysis of the messages encrypted with a Heer, M3 or M4 Enigma machine that was used by Germany in World War II.

A significant speed improvement is achieved by doing the computations on a GPU processor of a video card. For best results, Nvidia GTX 1070 or better is recommended.

The algorithm is similar to the one implemented in the enigma-suite software developed by Stefan Krah and used in the Enigma@Home project, with the improvements described here and the new partial exhaustion method described by Olaf Ostwald and Frode Weierud.

If compiled with a dynamically linked runtime library, the program requires installation of Visual C++ Redistributable for Visual Studio 2015

command-line options

The command line parameters are compatible with the enigma-suite program, however, some non-essential options are ignored, and some new options are supported.

usage:

enigma-cuda.exe <options> <trigram file name> <uni/bigram file name> <ciphertext file name>

options:

-h show help

-v show version number

-R resume operation using the state saved in the 00hc.resume file. The file is saved in the current working directory, so its location is sometimes unpredictable. This behavior is preserved for compatibility with enigma-suite. The format of the file is described in the enigma.txt document included with the enigma-suite source code. Note that this file does not save information about the -g, -e, -E and -s settings.

Edit: The -g option is now stored in the Resume file which breaks compatibility with the original enigma-suite format.

-o <file_name> save output to file

-M <model> Enigma model : H, M3 or M4. If not specified, the model is inferred from the keys

-f <key> first key. The format of the key is the same as in enigma-suite.

-t <key> last key. The program tries all keys in the range from first key to last key. If both keys are omitted, all possible keys are tried. Since computations are performed in blocks of 26^3 keys, the last three letters in the first and last keys are ignored and assumed to be AAA to ZZZ. The only exception to this is the case when the first and last keys are exactly the same, then only that specific key is tried.

-x turnover mode. Try the keys that result in a left hand wheel turnover within the message. The default behavior is the opposite, only the keys without a turnover are tried.

-a turnover mode. Try both keys with and without a left hand wheel turnover. The "duplicate" keys that produce the same plain text as some other key are still ignored in this mode

-n <count> the number of passes to make, 1 by default

-z <score> stop when this score is reached

-s <plugs> known plugboard connections. E.g., ABXY defines the A-B and X-Y plugs. The program will not touch these plugs during hill climbing. Useful when some plugboard connections are already known, e.g., from running a bombe

-e <letters> exhaustive search with a single fixed plug. E.g., ENRXSI tells the program to try the E-A..E-Z, N-A..N-Z, ... I-A..I-Z plugs. For 6 letters the total number of fixed plugs that will be tried is 141.

-E <letters> exhaustive search with multiple fixed plugs. Currently only two plugs at a time are allowed, E.g., with the EN argument the program will try 625 plug pairs, EANB to EZNY

-p start with a random swapping order. Withoug this option, the swapping order in the first pass is determined by the letter frequencies in the ciphertext

-g <scoring functions> use scoring functions: 0 = IC, 1 = unigrams, 2 = bigrams, 3 = trigrams. The default is 023.

For compatibility with enigma-suite, the program can use only the unigrams or bigrams file, but not both.

-d <device number> Use the specified GPU device. Device indices start with 0. If this switch is not present, the GPU with the highest number of multiprocessors is used.

-icuwrmk these options of enigma-suite are not supported by enigma-cuda, but their inclusion on the command line does not cause an error

feedback

Please send your feedback to

Alex Shovkoplyas VE3NEA