A typical data file often has associated "metadata" which is descriptive information about the file, represented in the form of a set of name-value pairs. Common metadata include creator's name, tools used to generate the file, file creation/update date, location of creation, editing history, etc. EXIF (images), RDF (web resources), DOI (digital documents) are some of popular metadata standards.
While metadata has its own merits in data management fields, it can actually affect your privacy [adversely][1]. EXIF data in photo images can reveal personally identifiable information such as your camera model, GPS coordinate of shooting, your favorite photo editor software, etc. Metadata in documents and spreadsheets contain author/affiliation information and other editing history. Not to be paranoid, but metadata gathering tools such as [metagoofil][2] are often exploited during information gathering stage as part of penetration testing.
For those of you who want to strip any personalizing metadata from any shared data, there are ways to remove metadata from data files. You can use existing document or image editor software which typically have built-in metadata editing capability. In this tutorial, let me introduce a nice standalone **metadata cleaner tool** which is developed for a single goal: **anonymize all metadata for your privacy**.
[MAT][3] (Metadata Anonymisation Toolkit) is a dedicated metadata cleaner written in Python. It was developed under the umbrella of the Tor project, and comes standard on [Tails][4], privacy-enhanced live OS.
Compared to other tools such as [exiftool][5] which can write to only a limited number of file types, MAT can eliminate metadata from all kinds of files: images (png, jpg), documents (odt, docx, pptx, xlsx, pdf), archives (tar, tar.bz2), audio (mp3, ogg, flac), etc.
### Install MAT on Linux ###
On Debian-based systems (Ubuntu or Linux Mint), MAT comes packaged, so installation is straightforward:
$ sudo apt-get install mat
On Fedora, MAT does not come as a pre-built package, so you need to build it from the source. Here is how I built MAT on Fedora (with some limited success; see the bottom of the tutorial):
If you don't supply any option with mat command, the default action is to remove metadata from files. If you want to keep a backup of original files during cleanup, use '-b' option. The following command cleans up all files, and stores original files as '*.bak" files.
Currently I have the following issue with a compiled version of MAT on Fedora. When I attempt to clean up archive/document files (e.g., *.gz, *.odt, *.docx) on Fedora, MAT fails with the following error. If you know how to fix this problem, let me know in the comment.
File "/usr/lib64/python2.7/zipfile.py", line 305, in __init__
raise ValueError('ZIP does not support timestamps before 1980')
ValueError: ZIP does not support timestamps before 1980
### Conclusion ###
MAT is a simple, yet extremely useful tool to prevent any inadvertent privacy leaks from metadata. Note that it is still your responsibility to anonymize file content, if necessary. All MAT does is to eliminate metadata associated with your files, but does nothing with the files themselves. In short, MAT can be a life saver as it can handle most common metadata removal, but you shouldn't rely solely on it to guarantee your privacy.
