masecret

masecret is a command to mask secret information in image files using OCR.

DISCLAIMER: There is no guarantee that all the secret information is successfully masked. You must make sure that all the secret information is masked.

Before:

After:

Prerequisite

• Python 3.3+
• Tesseract
  • Language data for OCR (can be specified with --lang, default is eng) must be available.

Installation

$ pip3 install masecret

You may need sudo.

masecret depends on pyocr and Pillow. If you fail to install Pillow, please see the installation instruction of Pillow.

Usage

Mask a single image file with a regular expression pattern that match AWS account number:

$ masecret -r '[-\d]{12,}' original.png -o masked.png

Mask multiple image files (output directory must exist):

$ masecret -r '[-\d]{12,}' original1.png original2.png ... -o masked_images/

Mask image files in-place with -i option:

$ masecret -i -r '[-\d]{12,}' original1.png original2.png ...

WARNING: No backup files will be created.

SECRETS.txt

When -r option is not provided, regular expressions are read from a file named SECRETS.txt in a current directory. Content of the file is regular expression patterns that match secret information you want to mask. You can include multiple patterns line by line.

Full Usage

usage:
    masecret [options] INPUT -o OUTPUT
    masecret [options] INPUT... -o OUTPUT
    masecret -i [options] INPUT...

Mask secret information in image files using OCR. Put regular expression
matches secret information into a file named SECRETS.txt or -r option.

positional arguments:
  INPUT                 input files

optional arguments:
  -h, --help            show this help message and exit
  -V, --version         show program's version number and exit
  -o OUTPUT, --output OUTPUT
                        output file or directory (default: None)
  -r REGEX, --regex REGEX
                        regular expression matches secret information
                        (default: None)
  -s SECRET_PATH, --secret SECRET_PATH
                        path to file containing regexes line by line that
                        match secret information (default: ./SECRETS.txt)
  -l LANG, --lang LANG  language for OCR, can be multiple languages joined by
                        + sign, e.g. eng+jpn (default: eng)
  -c COLOR, --color COLOR
                        color to fill secrets (default: #666)
  -i, --in-place        mask image files in-place. WARNING: No backup files
                        will be saved (default: False)
  --tesseract-params PARAMS
                        (Advanced Option) additional parameters passed to
                        tesseract (default: -psm 6 makebox)

Debug

If images are not masked as expected, the environment variable DEBUG will help you. If DEBUG is set, all the characters tesseract recognized are printed with position.

$ DEBUG=1 masecret original.png -o masked.png
Processing original.png...
. ((136, 90), (160, 114))
. ((176, 90), (200, 114))
. ((216, 90), (240, 114))
I ((292, 104), (304, 126))
I ((308, 104), (320, 126))
A ((326, 104), (340, 120))
W ((341, 104), (361, 120))
S ((362, 103), (375, 120))
M ((385, 104), (401, 120))
a ((404, 108), (415, 120))
n ((417, 108), (427, 120))
a ((430, 108), (440, 120))
g ((443, 108), (453, 125))
e ((456, 108), (467, 120))
m ((469, 108), (485, 120))
e ((488, 108), (499, 120))
n ((501, 108), (511, 120))
t ((513, 105), (519, 120))
C ((528, 103), (542, 120))
o ((545, 108), (556, 120))
n ((559, 108), (569, 120))
...

License

MIT License. See: LICENSE.

Packaging

(venv) $ pip install -r dev-requirements.txt
(venv) $ nosetests
(venv) $ pip install twine keyring
(venv) $ rm -rf dist
(venv) $ python setup.py sdist bdist_wheel
(venv) $ twine check dist/*
(venv) $ twine upload dist/*