timwhitez starred xortool

A tool to do some xor analysis:

guess the key length (based on count of equal chars)
guess the key (base on knowledge of most frequent char)

Notice: xortool is now only running on Python 3. The old Python 2 version is accessible at the py2 branch. The pip package has been updated.

Installation

For development or building this repository, poetry is needed.

poetry build
pip install dist/xortool*.whl

Usage

xortool
  A tool to do some xor analysis:
  - guess the key length (based on count of equal chars)
  - guess the key (base on knowledge of most frequent char)

Usage:
  xortool [-x] [-m MAX-LEN] [-f] [-t CHARSET] [FILE]
  xortool [-x] [-l LEN] [-c CHAR | -b | -o] [-f] [-t CHARSET] [-p PLAIN] [FILE]
  xortool [-x] [-m MAX-LEN| -l LEN] [-c CHAR | -b | -o] [-f] [-t CHARSET] [-p PLAIN] [FILE]
  xortool [-h | --help]
  xortool --version

Options:
  -x --hex                          input is hex-encoded str
  -l LEN, --key-length=LEN          length of the key
  -m MAX-LEN, --max-keylen=MAX-LEN  maximum key length to probe [default: 65]
  -c CHAR, --char=CHAR              most frequent char (one char or hex code)
  -b --brute-chars                  brute force all possible most frequent chars
  -o --brute-printable              same as -b but will only check printable chars
  -f --filter-output                filter outputs based on the charset
  -t CHARSET --text-charset=CHARSET target text character set [default: printable]
  -p PLAIN --known-plaintext=PLAIN  use known plaintext for decoding
  -h --help                         show this help

Notes:
  Text character set:
    * Pre-defined sets: printable, base32, base64
    * Custom sets:
      - a: lowercase chars
      - A: uppercase chars
      - 1: digits
      - !: special chars
      - *: printable chars

Examples:
  xortool file.bin
  xortool -l 11 -c 20 file.bin
  xortool -x -c ' ' file.hex
  xortool -b -f -l 23 -t base64 message.enc
  xortool -b -p "xctf{" message.enc

Example 1

# xor is xortool/xortool-xor
tests $ xor -f /bin/ls -s "secret_key" > binary_xored

tests $ xortool binary_xored
The most probable key lengths:
   2:   5.0%
   5:   8.7%
   8:   4.9%
  10:   15.4%
  12:   4.8%
  15:   8.5%
  18:   4.8%
  20:   15.1%
  25:   8.4%
  30:   14.9%
Key-length can be 5*n
Most possible char is needed to guess the key!

# 00 is the most frequent byte in binaries
tests $ xortool binary_xored -l 10 -c 00
...
1 possible key(s) of length 10:
secret_key

# decrypted ciphertexts are placed in ./xortool_out/Number_<key repr>
# ( have no better idea )
tests $ md5sum xortool_out/0_secret_key /bin/ls
29942e290876703169e1b614d0b4340a  xortool_out/0_secret_key
29942e290876703169e1b614d0b4340a  /bin/ls

The most common use is to pass just the encrypted file and the most frequent character (usually 00 for binaries and 20 for text files) - length will be automatically chosen:

tests $ xortool tool_xored -c 20
The most probable key lengths:
   2:   5.6%
   5:   7.8%
   8:   6.0%
  10:   11.7%
  12:   5.6%
  15:   7.6%
  20:   19.8%
  25:   7.8%
  28:   5.7%
  30:   11.4%
Key-length can be 5*n
1 possible key(s) of length 20:
an0ther s3cret \xdd key

Here, the key is longer then default 32 limit:

tests $ xortool ls_xored -c 00 -m 64
The most probable key lengths:
   3:   3.3%
   6:   3.3%
   9:   3.3%
  11:   7.0%
  22:   6.9%
  24:   3.3%
  27:   3.2%
  33:   18.4%
  44:   6.8%
  55:   6.7%
Key-length can be 3*n
1 possible key(s) of length 33:
really long s3cr3t k3y... PADDING

So, if automated decryption fails, you can calibrate:

(-m) max length to try longer keys
(-l) selected length to see some interesting keys
(-c) the most frequent char to produce right plaintext

Example 2

We are given a message in encoded in Base64 and XORed with an unknown key.

# xortool message.enc
The most probable key lengths:
   2:   12.3%
   4:   13.8%
   6:   10.5%
   8:   11.5%
  10:   8.6%
  12:   9.4%
  14:   7.1%
  16:   7.8%
  23:   10.4%
  46:   8.7%
Key-length can be 4*n
Most possible char is needed to guess the key!

We can now test the key lengths while filtering the outputs so that it only keeps the plaintexts holding the character set of Base64. After trying a few lengths, we come to the right one, which gives only 1 plaintext with a percentage of valid characters above the default threshold of 95%.

$ xortool message.enc -b -f -l 23 -t base64
256 possible key(s) of length 23:
\x01=\x121#"0\x17\x13\t\x7f ,&/\x12s\x114u\x170#
\x00<\x130"#1\x16\x12\x08~!-\'.\x13r\x105t\x161"
\x03?\x103! 2\x15\x11\x0b}".$-\x10q\x136w\x152!
\x02>\x112 !3\x14\x10\n|#/%,\x11p\x127v\x143
\x059\x165\'&4\x13\x17\r{$("+\x16w\x150q\x134\'
...
Found 1 plaintexts with 95.0%+ valid characters
See files filename-key.csv, filename-char_used-perc_valid.csv

By filtering the outputs on the character set of Base64, we directly keep the only solution.

Information

Author: hellman

License: MIT License