.gitlab-ci.yml

-image: debian
+variables:
+  CONTAINER_REGISTRY: $CI_REGISTRY/georg/mat2-ci-images
+  GIT_DEPTH: "5"
+  GIT_STRATEGY: clone
 
 stages:
   - linting
   - test
 
-bandit:
-  stage: linting
-  script:  # TODO: remove B405 and B314
-  - apt-get -qqy update
-  - apt-get -qqy install --no-install-recommends python3-bandit
-  - bandit ./mat2 --format txt --skip B101
-  - bandit -r ./nautilus/ --format txt --skip B101
-  - bandit -r ./libmat2 --format txt --skip B101,B404,B603,B405,B314
-
-pylint:
+.prepare_env: &prepare_env
+  before_script:  # This is needed to not run the testsuite as root
+    - useradd --home-dir ${CI_PROJECT_DIR} mat2
+    - chown -R mat2 .
+  
+linting:ruff:
+  image: $CONTAINER_REGISTRY:linting
   stage: linting
   script:
-  - apt-get -qqy update
-  - apt-get -qqy install --no-install-recommends pylint3 python3-mutagen python3-gi-cairo gir1.2-poppler-0.18 gir1.2-gdkpixbuf-2.0
-  - pylint3 --extension-pkg-whitelist=cairo,gi ./libmat2 ./mat2
-  # Once nautilus-python is in Debian, decomment it form the line below
-  - pylint3 --extension-pkg-whitelist=Nautilus,GObject,Gtk,Gio,GLib,gi ./nautilus/mat2.py
+    - apt update
+    - apt install -qqy --no-install-recommends python3-venv
+    - python3 -m venv venv
+    - source venv/bin/activate
+    - pip3 install ruff
+    - ruff check .
 
-pyflakes:
+linting:mypy:
+  image: $CONTAINER_REGISTRY:linting
   stage: linting
   script:
-  - apt-get -qqy update
-  - apt-get -qqy install --no-install-recommends pyflakes3
-  - pyflakes3 ./libmat2 ./mat2 ./tests/ ./nautilus
+    - mypy --ignore-missing-imports mat2 libmat2/*.py
 
-mypy:
-  stage: linting
+tests:archlinux:
+  image: $CONTAINER_REGISTRY:archlinux
+  stage: test
   script:
-  - apt-get -qqy update
-  - apt-get -qqy install --no-install-recommends python3-pip
-  - pip3 install mypy
-  - mypy mat2 libmat2/*.py --ignore-missing-imports
-  - mypy --ignore-missing-imports ./nautilus/mat2.py
-
+    - python3 -m unittest discover -v
+  
 tests:debian:
+  image: $CONTAINER_REGISTRY:debian
+  stage: test
+  <<: *prepare_env
+  script:
+    - apt-get -qqy purge bubblewrap
+    - su - mat2 -c "python3-coverage run --branch -m unittest discover -s tests/"
+    - su - mat2 -c "python3-coverage report --fail-under=95 -m --include 'libmat2/*'"
+
+tests:debian_with_bubblewrap:
+  image: $CONTAINER_REGISTRY:debian
   stage: test
+  allow_failure: true
+  <<: *prepare_env
   script:
-  - apt-get -qqy update
-  - apt-get -qqy install --no-install-recommends python3-mutagen python3-gi-cairo gir1.2-poppler-0.18 gir1.2-gdkpixbuf-2.0 libimage-exiftool-perl python3-coverage
-  - python3-coverage run --branch -m unittest discover -s tests/
-  - python3-coverage report -m --include 'libmat2/*'
+    - apt-get -qqy install bubblewrap
+    - python3 -m unittest discover -v
 
 tests:fedora:
-  image: fedora
+  image: $CONTAINER_REGISTRY:fedora
   stage: test
-  tags:
-    - whitewhale
   script:
-  - dnf install -y python3 python3-mutagen python3-gobject gdk-pixbuf2 poppler-glib gdk-pixbuf2 gdk-pixbuf2-modules cairo-gobject cairo python3-cairo perl-Image-ExifTool mailcap
-  - gdk-pixbuf-query-loaders-64 > /usr/lib64/gdk-pixbuf-2.0/2.10.0/loaders.cache
-  - python3 setup.py test
+    - python3 -m unittest discover -v
 
-tests:archlinux:
-  image: archlinux/base
+tests:gentoo:
+  image: $CONTAINER_REGISTRY:gentoo
+  stage: test
+  <<: *prepare_env
+  script:
+    - su - mat2 -c "python3 -m unittest discover -v"
+
+tests:python3.7:
+  image: $CONTAINER_REGISTRY:python3.7
+  stage: test
+  script:
+    - python3 -m unittest discover -v
+
+tests:python3.8:
+  image: $CONTAINER_REGISTRY:python3.8
+  stage: test
+  script:
+    - python3 -m unittest discover -v
+
+tests:python3.9:
+  image: $CONTAINER_REGISTRY:python3.9
+  stage: test
+  script:
+    - python3 -m unittest discover -v
+
+tests:python3.10:
+  image: $CONTAINER_REGISTRY:python3.10
+  stage: test
+  script:
+    - python3 -m unittest discover -v
+
+tests:python3.11:
+  image: $CONTAINER_REGISTRY:python3.11
+  stage: test
+  script:
+    - python3 -m unittest discover -v
+
+tests:python3.12:
+  image: $CONTAINER_REGISTRY:python3.12
+  stage: test
+  script:
+    - python3 -m unittest discover -v
+
+tests:python3.13:
+  image: $CONTAINER_REGISTRY:python3.13
+  stage: test
+  script:
+    - python3 -m unittest discover -v
+
+tests:python3.14:
+  image: $CONTAINER_REGISTRY:python3.14
   stage: test
-  tags:
-    - whitewhale
   script:
-  - pacman -Sy --noconfirm python-mutagen python-gobject gdk-pixbuf2 poppler-glib gdk-pixbuf2 python-cairo perl-image-exiftool python-setuptools mailcap
-  - python3 setup.py test
+    - python3 -m unittest discover -v

.pylintrc

@@ -6,11 +6,13 @@ max-locals=20
 disable=
     fixme,
     invalid-name,
+    duplicate-code,
     missing-docstring,
     protected-access,
-		abstract-method,
-		wrong-import-position,
-		catching-non-exception,
-		cell-var-from-loop,
-		locally-disabled,
-		invalid-sequence-index,  # pylint doesn't like things like `Tuple[int, bytes]` in type annotation
+    abstract-method,
+    wrong-import-position,
+    catching-non-exception,
+    cell-var-from-loop,
+    locally-disabled,
+		raise-missing-from,
+    invalid-sequence-index,  # pylint doesn't like things like `Tuple[int, bytes]` in type annotation

CHANGELOG.md

+# 0.13.5 - 2025-01-09
+- Keep orientation metadata on jpeg and tiff files
+- Improve cairo-related error/exceptions handling
+- Improve the logging
+- Improve the sandboxing
+- Improve Python3.12 support
+- Improve MSOffice documents handling
+
+# 0.13.4 - 2023-08-02
+
+- Add documentation about mat2 on OSX
+- Make use of python3.7 constructs to simplify code
+- Use moderner type annotations
+- Harden get_meta in archive.py against variants of CVE-2022-35410 
+- Improve MSOffice document support
+- Package the manpage on pypi 
+
+# 0.13.3 - 2023-02-23
+
+- Fix a decorator argument
+
+# 0.13.2 - 2023-01-28
+
+- Fix a crash on some python versions
+
+# 0.13.1 - 2023-01-07
+
+- Improve xlsx support
+- Remove the Nautilus extension
+
+# 0.13.0 - 2022-07-06
+
+- Fix an arbitrary file read (CVE-2022-35410)
+- Add support for heic files 
+
+# 0.12.4 - 2022-04-30
+
+- Fix possible errors/crashes when processing multiple files
+  via the command line interface
+- Use a fixed PDF version for the output
+- Improve compatibility with modern versions of rsvg
+- Improve the robustness of the command line interface with
+  regard to control characters
+
+# 0.12.3 - 2022-01-06
+
+- Implement code for internationalization
+- Keep individual files compression type in zip files
+- Increase the robustness of mat2 against weird/corrupted files
+- Fix the dolphin integration
+- Add a fuzzer
+
+# 0.12.2 - 2021-08-29
+
+- Add support for aiff files
+- Improve MS Office support
+- Improve compatibility with newer/older version of mat2's dependencies
+- Fix possible issues with the resolution of processed pdf
+
+# 0.12.1 - 2021-03-19
+
+- Improve epub support
+- Improve MS Office support
+
+# 0.12.0 - 2020-12-18
+
+- Improve significantly MS Office formats support
+- Fix some typos in the Nautilus extension
+- Improve reliability of the mp3, pdf and svg parsers
+- Improve compatibility with ffmpeg when sandboxing is used
+- Improve the dolphin extension usability
+- libmat2 now raises a ValueError on malformed files while trying to 
+  find the right parser, instead of returning None
+
+# 0.11.0 - 2020-03-29
+
+- Improve significantly MS Office formats support
+- Refactor how mat2 looks for executables
+
+# 0.10.1 - 2020-02-09
+
+- Improve the documentation and the manpage
+- Improve the robustness of css, html, png, gdk-based, exiftool-based parsers
+- Future-proof a bit the testsuite
+- Handle tiff files with a .tif extension
+- Improve the sandbox' usability
+- Add support for wav files
+
+# 0.10.0 - 2019-11-30
+
+- Make mat2 work on Python3.8
+- Minor improvement of ppt handling
+- Minor improvement of odt handling
+- Add an integration KDE's file manager: Dolphin
+- mat2 now copies file permissions on the cleaned files
+- Add a flag to disable sandboxing
+- Tighten a bit the sandboxing
+- Improve handling of MSOffice documents
+- Add support for inplace cleaning
+- Better handling of mutually-exclusive arguments in the command line
+- Add support for svg
+- Add support for ppm
+- Cleaned zip files are compressed by default
+- Minor performances improvement when dealing with images/video files
+- Better handling of optional dependencies
+
+# 0.9.0 - 2019-05-10
+
+- Add tar/tar.gz/tar.bz2/tar.zx archives support
+- Add support for xhtml files
+- Improve handling of read-only files
+- Improve a bit the command line's documentation
+- Fix a confusing error message
+- Add even more tests
+- Usuals internal cleanups/refactorings
+
+# 0.8.0 - 2019-02-28
+
+- Add support for epub files
+- Fix the setup.py file crashing on non-utf8 platforms
+- Improve css support
+- Improve html support
+
+# 0.7.0 - 2019-02-17
+
+- Add support for wmv files
+- Add support for gif files
+- Add support for html files
+- Sandbox external processes via bubblewrap
+- Simplify archive-based formats processing
+- The Nautilus extension now plays nicer with other extensions
+
+# 0.6.0 - 2018-11-10
+
+- Add lightweight cleaning for jpeg
+- Add support for zip files
+- Add support for mp4 files
+- Improve metadata extraction for archives
+- Improve robustness against corrupted embedded files
+- Fix a possible security issue on some terminals (control character
+	injection via --show)
+- Various internal cleanup/improvements
+
+# 0.5.0 - 2018-10-23
+
+- Video (.avi files for now) support, via FFmpeg, optionally
+- Lightweight cleaning for png and tiff files
+- Processing files starting with a dash is now quicker
+- Metadata are now displayed sorted
+- Recursive metadata support for FLAC files
+- Unsupported extensions aren't displayed in `./mat2 -l` anymore
+- Improve the display when no metadata are found
+- Update the logo according to the GNOME guidelines
+- The testsuite is now runnable on the installed version of mat2
+- Various internal cleanup/improvements
+
 # 0.4.0 - 2018-10-03
 
 - There is now a policy, for advanced users, to deal with unknown embedded fileformats
@@ -17,12 +173,12 @@
 
 # 0.3.1 - 2018-09-01
 
-- Document how to install MAT2 for various distributions
+- Document how to install mat2 for various distributions
 - Fix various typos in the documentation/comments
-- Add ArchLinux to the CI to ensure that MAT2 is running on it
+- Add ArchLinux to the CI to ensure that mat2 is running on it
 - Fix the handling of files with a name ending in `.JPG`
 - Improve the detection of unsupported extensions in upper-case
-- Streamline MAT2's logging
+- Streamline mat2's logging
 
 
 # 0.3.0 - 2018-08-03
@@ -42,14 +198,14 @@
 - Simplify various code-paths
 - Remove superfluous debug message
 - Remove the `--check` option that never was implemented anyway
-- Add a `-c` option to check for MAT2's dependencies
+- Add a `-c` option to check for mat2's dependencies
 
 
 # 0.1.3 - 2018-07-06
 
-- Improve MAT2 resilience against corrupted images
+- Improve mat2 resilience against corrupted images
 - Check that the minimal version of Poppler is available
-- Simplify how MAT2 deals with office files
+- Simplify how mat2 deals with office files
 - Improve cleaning of office files
 	- Thumbnails are removed
 	- Revisions are removed
@@ -61,8 +217,8 @@
 - Rename some files to ease the packaging
 - Add linters to the CI (mypy, bandit and pyflakes)
 - Prevent exitftool-related parameters injections
-- Improve MAT2's resilience against corrupted files
-- Make MAT2 work on fedora, thanks to @atenart
+- Improve mat2's resilience against corrupted files
+- Make mat2 work on fedora, thanks to @atenart
 - Tighten the threat model
 - Simplify and improve how office files are handled
 

CONTRIBUTING.md

-# Contributing to MAT2
+# Contributing to mat2
 
-The main repository for MAT2 is on [0xacab]( https://0xacab.org/jvoisin/mat2 ),
-with a mirror on [gitlab.com]( https://gitlab.com/jvoisin/mat2 ).
+The main repository for mat2 is on [0xacab]( https://0xacab.org/jvoisin/mat2 ),
+but you can send patches to jvoisin by [email](https://dustri.org/) if you prefer.
 
 Do feel free to pick up [an issue]( https://0xacab.org/jvoisin/mat2/issues )
-and to send a pull-request. Please do check that everything is fine by running the
-testsuite with `python3 -m unittest discover -v` before submitting one :)
+and to send a pull-request.
+
+Before sending the pull-request, please do check that everything is fine by
+running the full test suite in GitLab. To do that, after forking mat2 in GitLab,
+you need to go in Settings -> CI/CD -> Runner and there enable shared runners.
+
+Mat2 also has unit tests (that are also run in the full test suite). You can run
+them with `python3 -m unittest discover -v`.
 
 If you're fixing a bug or adding a new feature, please add tests accordingly,
 this will greatly improve the odds of your merge-request getting merged.
@@ -16,7 +22,7 @@ If you're adding a new fileformat, please add tests for:
 2. Cleaning metadata
 3. Raising `ValueError` upon a corrupted file
 
-Since MAT2 is written in Python3, please conform as much as possible to the
+Since mat2 is written in Python3, please conform as much as possible to the
 [pep8]( https://pep8.org/ ) style; except where it makes no sense of course.
 
 # Doing a release
@@ -24,14 +30,16 @@ Since MAT2 is written in Python3, please conform as much as possible to the
 1. Update the [changelog](https://0xacab.org/jvoisin/mat2/blob/master/CHANGELOG.md)
 2. Update the version in the [mat2](https://0xacab.org/jvoisin/mat2/blob/master/mat2) file
 3. Update the version in the [setup.py](https://0xacab.org/jvoisin/mat2/blob/master/setup.py) file
-4. Update the version and date in the [man page](https://0xacab.org/jvoisin/mat2/blob/master/doc/mat2.1)
-5. Commit the changelog, man page, mat2 and setup.py files
-6. Create a tag with `git tag -s $VERSION`
-7. Push the commit with `git push origin master`
-8. Push the tag with `git push --tags`
-9. Create the signed tarball with `git archive --format=tar.xz --prefix=mat-$VERSION/ $VERSION > mat-$VERSION.tar.xz`
-10. Sign the tarball with `gpg --armor --detach-sign mat-$VERSION.tar.xz`
-11. Upload the result on Gitlab's [tag page](https://0xacab.org/jvoisin/mat2/tags) and add the changelog there
-12. Announce the release on the [mailing list](https://mailman.boum.org/listinfo/mat-dev)
-13. Tell the [downstreams](https://0xacab.org/jvoisin/mat2/blob/master/INSTALL.md) about it
-14. Do the secret release dance
+4. Update the version in the [pyproject.toml](https://0xacab.org/jvoisin/mat2/blob/master/yproject.toml) file
+5. Update the version and date in the [man page](https://0xacab.org/jvoisin/mat2/blob/master/doc/mat2.1)
+6. Commit the modified files
+7. Create a tag with `git tag -s $VERSION`
+8. Push the commit with `git push origin master`
+9. Push the tag with `git push --tags`
+10. Download the gitlab archive of the release
+11. Diff it against the local copy
+12. If there is no difference, sign the archive with `gpg --armor --detach-sign mat2-$VERSION.tar.xz`
+13. Upload the signature on Gitlab's [tag page](https://0xacab.org/jvoisin/mat2/tags) and add the changelog there
+14. Announce the release on the [mailing list](https://mailman.boum.org/listinfo/mat-dev)
+15. Sign'n'upload the new version on pypi with `python3 setup.py sdist bdist_wheel` then `twine upload -s dist/*`
+16. Do the secret release dance

INSTALL.md

+# Python ecosystem
+
+If you feel like running arbitrary code downloaded over the
+internet (pypi doesn't support gpg signatures [anymore](https://github.com/pypa/python-packaging-user-guide/pull/466)),
+mat2 is [available on pypi](https://pypi.org/project/mat2/), and can be
+installed like this:
+
+```
+pip3 install mat2
+```
+
 # GNU/Linux
 
+## Optional dependencies
+
+When [bubblewrap](https://github.com/projectatomic/bubblewrap) is
+installed, mat2 uses it to sandbox any external processes it invokes.
+
+## Arch Linux
+
+Thanks to [kpcyrd](https://archlinux.org/packages/?maintainer=kpcyrd), there is an package available on
+[Arch linux's AUR](https://archlinux.org/packages/extra/any/mat2/).
+
+## Debian
+
+There is a package available in [Debian](https://packages.debian.org/search?keywords=mat2&searchon=names&section=all) and you can install mat2 with:
+
+```
+apt install mat2
+```
+
 ## Fedora
 
 Thanks to [atenart](https://ack.tf/), there is a package available on
 [Fedora's copr]( https://copr.fedorainfracloud.org/coprs/atenart/mat2/ ).
 
-We use copr (cool other packages repo) as the Mat2 Nautilus plugin depends on
-python3-nautilus, which isn't available yet in Fedora (but is distributed
-through this copr).
-
-First you need to enable Mat2's copr:
+First you need to enable mat2's copr:
 
 ```
 dnf -y copr enable atenart/mat2
 ```
 
-Then you can install both the Mat2 command and Nautilus extension:
+Then you can install mat2:
 
 ```
-dnf -y install mat2 mat2-nautilus
+dnf -y install mat2
 ```
 
-## Debian
+## Gentoo
 
-There is currently no package for Debian. If you want to help to make this
-happen, there is an [issue](https://0xacab.org/jvoisin/mat2/issues/16) open.
+mat2 is available in the [torbrowser overlay](https://github.com/MeisterP/torbrowser-overlay).
 
-But fear not, there is a way to install it *manually*:
 
-```
-# apt install python3-mutagen python3-gi-cairo gir1.2-gdkpixbuf-2.0 libimage-exiftool-perl gir1.2-glib-2.0 gir1.2-poppler-0.18
-$ git clone https://0xacab.org/jvoisin/mat2.git
-$ cd mat2
-$ ./mat2
-```
+# OSX
 
-and if you want to install the über-fancy Nautilus extension:
+## Homebrew
+
+mat2 is [available on homebrew](https://formulae.brew.sh/formula/mat2):
 
 ```
-# apt install gnome-common gtk-doc-tools libnautilus-extension-dev python-gi-dev
-$ git clone https://github.com/GNOME/nautilus-python
-$ cd nautilus-python
-$ PYTHON=/usr/bin/python3 ./autogen.sh
-$ make
-# make install
-$ mkdir -p ~/.local/share/nautilus-python/extensions/
-$ cp ../nautilus/mat2.py ~/.local/share/nautilus-python/extensions/
-$ PYTHONPATH=/home/$USER/mat2 PYTHON=/usr/bin/python3 nautilus
+brew install mat2
 ```
 
-## Arch Linux
-
-Thanks to [Francois_B](https://www.sciunto.org/), there is an package available on
-[Arch linux's AUR](https://aur.archlinux.org/packages/mat2/).
+## MacPorts
 
-## Gentoo
+mat2 is [available on MacPorts](https://ports.macports.org/port/mat2/):
 
-MAT2 is available in the [torbrowser overlay](https://github.com/MeisterP/torbrowser-overlay).
+```
+port install mat2
+```

README.md

 ```
  _____ _____ _____ ___
 |     |  _  |_   _|_  |  Keep your data,
-| | | |     | | | |  _|     trash your meta!
-|_|_|_|__|__| |_| |___|
+| | | | |_| | | | |  _|     trash your meta!
+|_|_|_|_| |_| |_| |___|
 
 ```
 
-This software is currently in **beta**, please don't use it for anything
-critical.
-
 # Metadata and privacy
 
 Metadata consist of information that characterizes data.
@@ -20,20 +17,37 @@ Metadata within a file can tell a lot about you.
 Cameras record data about when a picture was taken and what
 camera was used. Office documents like PDF or Office automatically adds
 author and company information to documents and spreadsheets.
-Maybe you don't want to disclose those information on the web.
+Maybe you don't want to disclose those information.
 
-This is precisely the job of MAT2: getting rid, as much as possible, of
+This is precisely the job of mat2: getting rid, as much as possible, of
 metadata.
 
+mat2 provides:
+- a library called `libmat2`;
+- a command line tool called `mat2`,
+- a service menu for Dolphin, KDE's default file manager
+
+If you prefer a regular graphical user interface, you might be interested in
+[Metadata Cleaner](https://metadatacleaner.romainvigier.fr/), which is using
+`mat2` under the hood.
+
 # Requirements
 
 - `python3-mutagen` for audio support
 - `python3-gi-cairo` and `gir1.2-poppler-0.18` for PDF support
 - `gir1.2-gdkpixbuf-2.0` for images support
+- `gir1.2-rsvg-2.0` for svg support
+- `FFmpeg`, optionally, for video support
 - `libimage-exiftool-perl` for everything else
+- `bubblewrap`, optionally, for sandboxing
+
+Please note that mat2 requires at least Python3.5.
 
-Please note that MAT2 requires at least Python3.5, meaning that it
-doesn't run on [Debian Jessie](https://packages.debian.org/jessie/python3),
+# Requirements setup on macOS (OS X) using [Homebrew](https://brew.sh/)
+
+```bash
+brew install exiftool cairo pygobject3 poppler gdk-pixbuf librsvg ffmpeg
+```
 
 # Running the test suite
 
@@ -41,11 +55,18 @@ doesn't run on [Debian Jessie](https://packages.debian.org/jessie/python3),
 $ python3 -m unittest discover -v
 ```
 
-# How to use MAT2
+And if you want to see the coverage:
 
 ```bash
-usage: mat2 [-h] [-v] [-l] [--check-dependencies] [-V]
-            [--unknown-members policy] [-s | -L]
+$ python3-coverage run --branch -m unittest discover -s tests/
+$ python3-coverage report --include -m --include /libmat2/*'
+```
+
+# How to use mat2
+
+```
+usage: mat2 [-h] [-V] [--unknown-members policy] [--inplace] [--no-sandbox]
+            [-v] [-l] [--check-dependencies] [-L | -s]
             [files [files ...]]
 
 Metadata anonymisation toolkit 2
@@ -55,32 +76,64 @@ positional arguments:
 
 optional arguments:
   -h, --help            show this help message and exit
-  -v, --version         show program's version number and exit
-  -l, --list            list all supported fileformats
-  --check-dependencies  check if MAT2 has all the dependencies it needs
   -V, --verbose         show more verbose status information
   --unknown-members policy
                         how to handle unknown members of archive-style files
-                        (policy should be one of: abort, omit, keep)
-  -s, --show            list harmful metadata detectable by MAT2 without
-                        removing them
+                        (policy should be one of: abort, omit, keep) [Default:
+                        abort]
+  --inplace             clean in place, without backup
+  --no-sandbox          Disable bubblewrap's sandboxing
+  -v, --version         show program's version number and exit
+  -l, --list            list all supported fileformats
+  --check-dependencies  check if mat2 has all the dependencies it needs
   -L, --lightweight     remove SOME metadata
+  -s, --show            list harmful metadata detectable by mat2 without
+                        removing them
 ```
 
-Note that MAT2 **will not** clean files in-place, but will produce, for
+Note that mat2 **will not** clean files in-place, but will produce, for
 example, with a file named "myfile.png" a cleaned version named
 "myfile.cleaned.png".
 
+## Web interface
+
+It's possible to run mat2 as a web service, via
+[mat2-web](https://0xacab.org/jvoisin/mat2-web).
+
+If you're using WordPress, you might be interested in [wp-mat](https://git.autistici.org/noblogs/wp-mat)
+and [wp-mat-server](https://git.autistici.org/noblogs/wp-mat-server).
+
+## Desktop GUI
+
+For GNU/Linux desktops, it's possible to use the
+[Metadata Cleaner](https://gitlab.com/rmnvgr/metadata-cleaner) GTK application.
+
+# Supported formats
+
+The following formats are supported: avi, bmp, css, epub/ncx, flac, gif, jpeg,
+m4a/mp2/mp3/…, mp4, odc/odf/odg/odi/odp/ods/odt/…, off/opus/oga/spx/…, pdf,
+png, ppm, pptx/xlsx/docx/…, svg/svgz/…, tar/tar.gz/tar.bz2/tar.xz/…, tiff,
+torrent, wav, wmv, zip, …
+  
 # Notes about detecting metadata
 
-While MAT2 is doing its very best to display metadata when the `--show` flag is
-passed, it doesn't mean that a file is clean from any metadata if MAT2 doesn't
+While mat2 is doing its very best to display metadata when the `--show` flag is
+passed, it doesn't mean that a file is clean from any metadata if mat2 doesn't
 show any. There is no reliable way to detect every single possible metadata for
 complex file formats.
 
 This is why you shouldn't rely on metadata's presence to decide if your file must
 be cleaned or not.
 
+# Notes about the lightweight mode
+
+By default, mat2 might alter a bit the data of your files, in order to remove
+as much metadata as possible. For example, texts in PDF might not be selectable anymore,
+compressed images might get compressed again, …
+Since some users might be willing to trade some metadata's presence in exchange
+of the guarantee that mat2 won't modify the data of their files, there is the
+`-L` flag that precisely does that.
+
 # Related software
 
 - The first iteration of [MAT](https://mat.boum.org)
@@ -91,15 +144,21 @@ be cleaned or not.
 	watermarks from PDF.
 - [Scrambled Exif](https://f-droid.org/packages/com.jarsilio.android.scrambledeggsif/),
 	an open-source Android application to remove metadata from pictures.
+- [Dangerzone](https://dangerzone.rocks/), designed to sanitize harmful documents
+  into harmless ones.
 
 # Contact
 
 If possible, use the [issues system](https://0xacab.org/jvoisin/mat2/issues)
-or the [mailing list](https://mailman.boum.org/listinfo/mat-dev)
+or the [mailing list](https://www.autistici.org/mailman/listinfo/mat-dev)
 Should a more private contact be needed (eg. for reporting security issues),
 you can email Julien (jvoisin) Voisin at `julien.voisin+mat2@dustri.org`,
 using the gpg key `9FCDEE9E1A381F311EA62A7404D041E8171901CC`.
 
+# Donations
+
+If you want to donate some money, please give it to [Tails]( https://tails.boum.org/donate/?r=contribute ).
+
 # License
 
 This program is free software: you can redistribute it and/or modify
@@ -115,16 +174,20 @@ GNU General Public License for more details.
 You should have received a copy of the GNU Lesser General Public License
 along with this program.  If not, see <http://www.gnu.org/licenses/>.
 
-Copyright 2018 Julien (jvoisin) Voisin <julien.voisin+mat2@dustri.org>
-Copyright 2016 Marie Rose for MAT2's logo
+Copyright 2018 Julien (jvoisin) Voisin <julien.voisin+mat2@dustri.org>  
+Copyright 2016 Marie-Rose for mat2's logo
+
+The `tests/data/dirty_with_nsid.docx` file is licensed under GPLv3,
+and was borrowed from the Calibre project: https://calibre-ebook.com/downloads/demos/demo.docx
+
+The `narrated_powerpoint_presentation.pptx` file is in the public domain.
 
 # Thanks
 
-MAT2 wouldn't exist without:
+mat2 wouldn't exist without:
 
 - the [Google Summer of Code](https://summerofcode.withgoogle.com/);
 - the fine people from [Tails]( https://tails.boum.org);
 - friends
 
 Many thanks to them!
-

doc/comparison_to_others.md

+# Exiftool
+
+mat2 is in fact using exiftool to extract metadata from files,
+but not to remove them. The previous iteration of mat2, MAT,
+was using exiftool to remove metadata, which lead to several cases where
+they weren't correctly removed, if at all.
+For example, [Exiftool's documentation](https://www.sno.phy.queensu.ca/~phil/exiftool/TagNames/PDF.html)
+states the following with regard to PDF:
+
+> All metadata edits are reversible. While this would normally be considered an
+advantage, it is a potential security problem because old information is never
+actually deleted from the file.
+
+To remove metadata, mat2 usually re-render the file completely, eliminating
+all possible original metadata. See the `implementation_notes.md` file for
+details.
+
+
+# jpegoptim, optipng, …
+
+While designed to reduce as much as possible the size of pictures,
+those software can be used to remove metadata. They usually have very good
+support for a single picture format, and can be used in place of mat2 for them.
+
+
+# PDF Redact Tools
+
+[PDF Redact Tools](https://github.com/firstlookmedia/pdf-redact-tools) is
+a software developed by the people from [First Look
+Media](https://firstlook.media/), the entity behind, amongst other things, 
+[The Intercept](https://theintercept.com/).
+
+The tool uses roughly the same approach than mat2 to deal with PDF,
+which is unfortunately the only fileformat that it does support.
+It's interesting to note that it has counter-measures against
+[yellow dots](https://en.wikipedia.org/wiki/Machine_Identification_Code),
+a capacity that mat2 [doesn't possess yet](https://0xacab.org/jvoisin/mat2/issues/43).
+
+
+# Exiv2
+
+[Exiv2](https://www.exiv2.org/) was considered for mat2,
+but it currently [misses a lot of metadata](https://0xacab.org/jvoisin/mat2/issues/85)
+
+
+# Others non open source software/online service
+
+There are a lot of closed-source software and online service claiming to remove
+metadata from your files, but since there is no way to actually verify that
+they're effectively removing them, let alone adding unique markers, they
+shouldn't be used.

doc/implementation_notes.md

@@ -4,7 +4,7 @@ Implementation notes
 Lightweight cleaning mode
 -------------------------
 
-Due to *popular* request, MAT2 is providing a *lightweight* cleaning mode,
+Due to *popular* request, mat2 is providing a *lightweight* cleaning mode,
 that only cleans the superficial metadata of your file, but not
 the ones that might be in **embedded** resources. Like for example,
 images in a PDF or an office document.
@@ -12,33 +12,49 @@ images in a PDF or an office document.
 Revisions handling
 ------------------
 
-Revisions are handled according to the principle of least astonishment: they are entirely removed.
+Revisions are handled according to the principle of least astonishment: they
+are entirely removed.
 
-- Either the users aren't aware of the revisions, are thus they should be deleted. For example journalists that are editing a document to erase mentions sources mentions.
+- Either the users aren't aware of the revisions, are thus they should be
+	deleted. For example journalists that are editing a document to erase
+	mentions sources mentions.
 
-- Or they are aware of it, and will likely not expect MAT2 to be able to keep the revisions, that are basically traces about how, when and who edited the document.
+- Or they are aware of it, and will likely not expect mat2 to be able to keep
+	the revisions, that are basically traces about how, when and who edited the
+	document.
 
 
 Race conditions
 ---------------
 
-MAT2 does its very best to avoid crashing at runtime. This is why it's checking
-if the file is valid __at parser creation__. MAT2 doesn't take any measure to
+mat2 does its very best to avoid crashing at runtime. This is why it's checking
+if the file is valid __at parser creation__. mat2 doesn't take any measure to
 ensure that the file is not changed between the time the parser is
 instantiated, and the call to clean or show the metadata.
 
 Symlink attacks
 ---------------
 
-MAT2 output predictable filenames (like yourfile.jpg.cleaned).
+mat2 output predictable filenames (like yourfile.jpg.cleaned).
 This may lead to symlink attack. Please check if you OS prevent
 against them
 
 Archives handling
 -----------------
 
-MAT2 doesn't support archives yet, because we haven't found an usable way to ask the user
-what to do when a non-supported files are encountered.
+By default, when cleaning a non-support file format in an archive,
+mat2 will abort with a detailed error message.
+While strongly discouraged, it's possible to override this behaviour to force
+the exclusion, or inclusion of unknown files into the cleaned archive.
+
+While Python's [zipfile](https://docs.python.org/3/library/zipfile.html) module
+provides *safe* way to extract members of a zip archive, the 
+[tarfile](https://docs.python.org/3/library/tarfile.html) one doesn't,
+meaning that it's up to mat2 to implement safety checks. Currently,
+it defends against path-traversal, both relative and absolute,
+symlink-related attacks, setuid/setgid attacks, duplicate members, block and
+char devices, … but there might still be dragons lurking there.
+
 
 PDF handling
 ------------
@@ -49,10 +65,10 @@ didn't remove any *deep metadata*, like the ones in embedded pictures. This was
 on of the reason MAT was abandoned: the absence of satisfying solution to
 handle PDF. But apparently, people are ok with [pdf redact
 tools](https://github.com/firstlookmedia/pdf-redact-tools), that simply
-transform the PDF into images. So this is what's MAT2 is doing too.
+transform the PDF into images. So this is what's mat2 is doing too.
 
 Of course, it would be possible to detect images in PDf file, and process them
-with MAT2, but since a PDF can contain a lot of things, like images, videos,
+with mat2, but since a PDF can contain a lot of things, like images, videos,
 javascript, pdf, blobs, … this is the easiest and safest way to clean them.
 
 Images handling
@@ -65,7 +81,7 @@ XML attacks
 -----------
 
 Since our threat model conveniently excludes files crafted to specifically
-bypass MAT2, fileformats containing harmful XML are out of our scope.
-But since MAT2 is using [etree](https://docs.python.org/3/library/xml.html#xml-vulnerabilities)
+bypass mat2, fileformats containing harmful XML are out of our scope.
+But since mat2 is using [etree](https://docs.python.org/3/library/xml.html#xml-vulnerabilities)
 to process XML, it's "only" vulnerable to DoS, and not memory corruption:
 odds are that the user will notice that the cleaning didn't succeed.

doc/mat2.1

-.TH MAT2 "1" "October 2018" "MAT2 0.4.0" "User Commands"
+.TH mat2 "1" "January 2025" "mat2 0.13.5" "User Commands"
 
 .SH NAME
 mat2 \- the metadata anonymisation toolkit 2
@@ -8,7 +8,7 @@ mat2 \- the metadata anonymisation toolkit 2
 
 .SH DESCRIPTION
 .B mat2
-removes metadata from various fileformats. It supports a wide variety of file 
+removes metadata from various fileformats. It supports a wide variety of file
 formats, audio, office, images, …
 
 Careful, mat2 does not clean files in-place, instead, it will produce a file with the word
@@ -32,7 +32,7 @@ show program's version number and exit
 list all supported fileformats
 .TP
 \fB\-\-check\-dependencies\fR
-check if MAT2 has all the dependencies it needs
+check if mat2 has all the dependencies it needs
 .TP
 \fB\-V\fR, \fB\-\-verbose\fR
 show more verbose status information
@@ -41,11 +41,16 @@ show more verbose status information
 how to handle unknown members of archive-style files (policy should be one of: abort, omit, keep)
 .TP
 \fB\-s\fR, \fB\-\-show\fR
-list harmful metadata detectable by MAT2 without
-removing them
+list harmful metadata detectable by mat2 without removing them
 .TP
 \fB\-L\fR, \fB\-\-lightweight\fR
 remove SOME metadata
+.TP
+\fB\--no-sandbox\fR
+disable bubblewrap's sandboxing
+.TP
+\fB\--inplace\fR
+clean in place, without backup
 
 .SH EXAMPLES
 To remove all the metadata from a PDF file:
@@ -57,6 +62,24 @@ mat2 ./myfile.pdf
 .fi
 .PP
 
+.SH NOTES ABOUT METADATA
+
+While mat2 is doing its very best to display metadata when the --show flag is
+passed, it doesn't mean that a file is clean from any metadata if mat2 doesn't
+show any. There is no reliable way to detect every single possible metadata for
+complex file formats.
+.PP
+This is why you shouldn't rely on metadata's presence to decide if your file must
+be cleaned or not.
+.PP
+Moreover, mat2 goes to great lengths to make sure that as much metadata as
+possible are removed. This might sometimes result in a loss of quality of the
+processed files. For example, textual based pdf file converted into image based
+one means that it'll be no longer possible to select text in them. If you're
+experiencing this, you might want to give the lightweight cleaning mode a try,
+but keep in mind by doing so, some metadata \fBwon't be cleaned\fR.
+
+
 .SH BUGS
 
 While mat2 does its very best to remove every single metadata,

doc/threat_model.md

@@ -3,7 +3,7 @@ Threat Model
 
 The Metadata Anonymisation Toolkit 2 adversary has a number
 of goals, capabilities, and counter-attack types that can be
-used to guide us towards a set of requirements for the MAT2.
+used to guide us towards a set of requirements for the mat2.
 
 This is an overhaul of MAT's (the first iteration of the software) one.
 
@@ -53,7 +53,7 @@ Adversary
     user. This is the strongest position for the adversary to
     have. In this case, the adversary is capable of inserting
     arbitrary, custom watermarks specifically for tracking
-    the user. In general, MAT2 cannot defend against this
+    the user. In general, mat2 cannot defend against this
     adversary, but we list it for completeness' sake.
 
     - The adversary created the document for a group of users.
@@ -65,7 +65,7 @@ Adversary
 		- The adversary did not create the document, the weakest
 		position for the adversary to have. The file format is
 		(most of the time) standard, nothing custom is added:
-		MAT2 must be able to remove all metadata from the file.
+		mat2 must be able to remove all metadata from the file.
 
 
 Requirements
@@ -73,28 +73,28 @@ Requirements
 
 * Processing
 
-    - MAT2 *should* avoid interactions with information.
+    - mat2 *should* avoid interactions with information.
     Its goal is to remove metadata, and the user is solely
     responsible for the information of the file.
 
-    - MAT2 *must* warn when encountering an unknown
-    format. For example, in a zipfile, if MAT encounters an
+    - mat2 *must* warn when encountering an unknown
+    format. For example, in a zipfile, if mat2 encounters an
     unknown format, it should warn the user, and ask if the
     file should be added to the anonymised archive that is
     produced.
 
-    - MAT2 *must* not add metadata, since its purpose is to
+    - mat2 *must* not add metadata, since its purpose is to
     anonymise files: every added items of metadata decreases
     anonymity.
 
-    - MAT2 *should* handle unknown/hidden metadata fields,
+    - mat2 *should* handle unknown/hidden metadata fields,
     like proprietary extensions of open formats.
 
-		- MAT2 *must not* fail silently. Upon failure,
-		MAT2 *must not* modify the file in any way.
+		- mat2 *must not* fail silently. Upon failure,
+		mat2 *must not* modify the file in any way.
 
-		- MAT2 *might* leak the fact that MAT2 was used on the file,
+		- mat2 *might* leak the fact that mat2 was used on the file,
 		since it might be uncommon for some file formats to come
 		without any kind of metadata, an adversary might suspect that
-		the user used MAT2 on certain files.
+		the user used mat2 on certain files.
 

dolphin/README.md

+Dolphin integration
+===================
+
+Thanks to [Miguel Marco](https://riemann.unizar.es/~mmarco/), here is an neat
+integration for [Dolphin](https://kde.org/applications/system/org.kde.dolphin),
+the KDE file manager:
+
+1. Add the `mat2.desktop` file either in
+	- `/usr/share/kservices5/ServiceMenus/` to install it globally
+	- `~/.local/share/kservices5/ServiceMenus/` for a specific user
+2. Run `kbuildsycoca5` to update the corresponding database
+3. Enjoy your new contextual menu to remove metadata from your files!
+
+

dolphin/mat2.desktop

+[Desktop Entry]
+X-KDE-ServiceTypes=KonqPopupMenu/Plugin
+MimeType=application/pdf;application/vnd.oasis.opendocument.chart;application/vnd.oasis.opendocument.formula;application/vnd.oasis.opendocument.graphics;application/vnd.oasis.opendocument.image;application/vnd.oasis.opendocument.presentation;application/vnd.oasis.opendocument.spreadsheet;application/vnd.oasis.opendocument.text;application/vnd.openxmlformats-officedocument.presentationml.presentation;application/vnd.openxmlformats-officedocument.spreadsheetml.sheet;application/vnd.openxmlformats-officedocument.wordprocessingml.document;application/x-bittorrent;application/zip;audio/flac;audio/mpeg;audio/ogg;audio/x-flac;image/jpeg;image/png;image/tiff;image/x-ms-bmp;text/plain;video/mp4;video/x-msvideo;
+Actions=cleanMetadata;
+Type=Service
+
+[Desktop Action cleanMetadata]
+Name=Clean metadata
+Name[de]=Metadaten löschen
+Name[es]=Limpiar metadatos
+Icon=/usr/share/icons/hicolor/scalable/apps/mat2.svg
+Exec=kdialog --yesno  "$( mat2 -s %F )" --title "Clean Metadata?" && mat2 %U
+Exec[de]=kdialog --yesno  "$( mat2 -s %F )" --title "Metadaten löschen?" && mat2 %U

libmat2/__init__.py

 #!/usr/bin/env python3
 
-import os
-import collections
 import enum
 import importlib
-from typing import Dict, Optional
+from typing import Dict
 
-# make pyflakes happy
-assert Dict
+from . import exiftool, video
 
 # A set of extension that aren't supported, despite matching a supported mimetype
 UNSUPPORTED_EXTENSIONS = {
@@ -28,42 +25,70 @@ UNSUPPORTED_EXTENSIONS = {
     }
 
 DEPENDENCIES = {
-    'cairo': 'Cairo',
-    'gi': 'PyGobject',
-    'gi.repository.GdkPixbuf': 'GdkPixbuf from PyGobject',
-    'gi.repository.Poppler': 'Poppler from PyGobject',
-    'gi.repository.GLib': 'GLib from PyGobject',
-    'mutagen': 'Mutagen',
-    }
-
-def _get_exiftool_path() -> Optional[str]:  # pragma: no cover
-    exiftool_path = '/usr/bin/exiftool'
-    if os.path.isfile(exiftool_path):
-        if os.access(exiftool_path, os.X_OK):
-            return exiftool_path
-
-    # ArchLinux
-    exiftool_path = '/usr/bin/vendor_perl/exiftool'
-    if os.path.isfile(exiftool_path):
-        if os.access(exiftool_path, os.X_OK):
-            return exiftool_path
+    'Cairo': {
+        'module': 'cairo',
+        'required': True,
+    },
+    'PyGobject': {
+        'module': 'gi',
+        'required': True,
+    },
+    'GdkPixbuf from PyGobject': {
+        'module': 'gi.repository.GdkPixbuf',
+        'required': True,
+    },
+    'Poppler from PyGobject': {
+        'module': 'gi.repository.Poppler',
+        'required': True,
+    },
+    'GLib from PyGobject': {
+        'module': 'gi.repository.GLib',
+        'required': True,
+    },
+    'Mutagen': {
+        'module': 'mutagen',
+        'required': True,
+    },
+}
 
-    return None
+CMD_DEPENDENCIES = {
+    'Exiftool': {
+        'cmd': exiftool._get_exiftool_path,
+        'required': False,
+    },
+    'Ffmpeg': {
+        'cmd': video._get_ffmpeg_path,
+        'required': False,
+    },
+}
 
-def check_dependencies() -> dict:
-    ret = collections.defaultdict(bool)  # type: Dict[str, bool]
 
-    ret['Exiftool'] = True if _get_exiftool_path() else False
+def check_dependencies() -> Dict[str, Dict[str, bool]]:
+    ret: Dict[str, Dict] = dict()
 
     for key, value in DEPENDENCIES.items():
-        ret[value] = True
+        ret[key] = {
+            'found': True,
+            'required': value['required'],
+        }
         try:
-            importlib.import_module(key)
+            importlib.import_module(value['module'])  # type: ignore
         except ImportError:  # pragma: no cover
-            ret[value] = False  # pragma: no cover
+            ret[key]['found'] = False
+
+    for k, v in CMD_DEPENDENCIES.items():
+        ret[k] = {
+            'found': True,
+            'required': v['required'],
+        }
+        try:
+            v['cmd']()  # type: ignore
+        except RuntimeError:  # pragma: no cover
+            ret[k]['found'] = False
 
     return ret
 
+
 @enum.unique
 class UnknownMemberPolicy(enum.Enum):
     ABORT = 'abort'

libmat2/abstract.py

 import abc
 import os
-from typing import Set, Dict
-
-assert Set  # make pyflakes happy
+import re
+from typing import Union, Set, Dict
 
 
 class AbstractParser(abc.ABC):
     """ This is the base class of every parser.
-    It might yield `ValueError` on instantiation on invalid files.
+    It might yield `ValueError` on instantiation on invalid files,
+    and `RuntimeError` when something went wrong in `remove_all`.
     """
-    meta_list = set()  # type: Set[str]
-    mimetypes = set()  # type: Set[str]
+    meta_list: Set[str] = set()
+    mimetypes: Set[str] = set()
 
     def __init__(self, filename: str) -> None:
         """
         :raises ValueError: Raised upon an invalid file
         """
+        if re.search('^[a-z0-9./]', filename) is None:
+            # Some parsers are calling external binaries,
+            # this prevents shell command injections
+            filename = os.path.join('.', filename)
+
         self.filename = filename
         fname, extension = os.path.splitext(filename)
+
+        # Special case for tar.gz, tar.bz2, … files
+        if fname.endswith('.tar') and len(fname) > 4:
+            fname, extension = fname[:-4], '.tar' + extension
+
         self.output_filename = fname + '.cleaned' + extension
+        self.lightweight_cleaning = False
+        self.sandbox = True
 
     @abc.abstractmethod
-    def get_meta(self) -> Dict[str, str]:
-        pass  # pragma: no cover
+    def get_meta(self) -> Dict[str, Union[str, Dict]]:
+        """Return all the metadata of the current file
+
+        :raises RuntimeError: Raised if the cleaning process went wrong.
+        """
 
     @abc.abstractmethod
     def remove_all(self) -> bool:
-        pass  # pragma: no cover
+        """
+        Remove all the metadata of the current file
 
-    def remove_all_lightweight(self) -> bool:
-        """ This method removes _SOME_ metadata.
-        It might be useful to implement it for fileformats that do
-        not support non-destructive cleaning.
+        :raises RuntimeError: Raised if the cleaning process went wrong.
         """
-        return self.remove_all()

libmat2/audio.py

+import mimetypes
+import os
 import shutil
+import tempfile
+from typing import Union, Dict
 
 import mutagen
 
-from . import abstract
+from . import abstract, parser_factory, video
 
 
 class MutagenParser(abstract.AbstractParser):
     def __init__(self, filename):
         super().__init__(filename)
         try:
-            mutagen.File(self.filename)
+            if mutagen.File(self.filename) is None:
+                raise ValueError
         except mutagen.MutagenError:
             raise ValueError
 
-    def get_meta(self):
+    def get_meta(self) -> Dict[str, Union[str, Dict]]:
         f = mutagen.File(self.filename)
         if f.tags:
-            return {k:', '.join(v) for k, v in f.tags.items()}
+            return {k: ', '.join(map(str, v)) for k, v in f.tags.items()}
         return {}
 
-    def remove_all(self):
+    def remove_all(self) -> bool:
         shutil.copy(self.filename, self.output_filename)
         f = mutagen.File(self.output_filename)
-        f.delete()
-        f.save()
+        try:
+            f.delete()
+            f.save()
+        except mutagen.MutagenError:
+            raise ValueError
         return True
 
 
 class MP3Parser(MutagenParser):
     mimetypes = {'audio/mpeg', }
 
-    def get_meta(self):
-        metadata = {}
+    def get_meta(self) -> Dict[str, Union[str, Dict]]:
+        metadata: Dict[str, Union[str, Dict]] = dict()
         meta = mutagen.File(self.filename).tags
+        if not meta:
+            return metadata
         for key in meta:
+            if isinstance(key, tuple):
+                metadata[key[0]] = key[1]
+                continue
+            if not hasattr(meta[key], 'text'):  # pragma: no cover
+                continue
             metadata[key.rstrip(' \t\r\n\0')] = ', '.join(map(str, meta[key].text))
         return metadata
 
@@ -44,3 +59,56 @@ class OGGParser(MutagenParser):
 
 class FLACParser(MutagenParser):
     mimetypes = {'audio/flac', 'audio/x-flac'}
+
+    def remove_all(self) -> bool:
+        shutil.copy(self.filename, self.output_filename)
+        f = mutagen.File(self.output_filename)
+        f.clear_pictures()
+        f.delete()
+        f.save(deleteid3=True)
+        return True
+
+    def get_meta(self) -> Dict[str, Union[str, Dict]]:
+        meta = super().get_meta()
+        for num, picture in enumerate(mutagen.File(self.filename).pictures):
+            name = picture.desc if picture.desc else 'Cover %d' % num
+            extension = mimetypes.guess_extension(picture.mime)
+            if extension is None: #  pragma: no cover
+                meta[name] = 'harmful data'
+                continue
+
+            _, fname = tempfile.mkstemp()
+            fname = fname + extension
+            with open(fname, 'wb') as f:
+                f.write(picture.data)
+            p, _ = parser_factory.get_parser(fname)  # type: ignore
+            if p is None:
+                raise ValueError
+            p.sandbox = self.sandbox
+            # Mypy chokes on ternaries :/
+            meta[name] = p.get_meta() if p else 'harmful data'  # type: ignore
+            os.remove(fname)
+        return meta
+
+
+class WAVParser(video.AbstractFFmpegParser):
+    mimetypes = {'audio/x-wav', }
+    meta_allowlist = {'AvgBytesPerSec', 'BitsPerSample', 'Directory',
+                      'Duration', 'Encoding', 'ExifToolVersion',
+                      'FileAccessDate', 'FileInodeChangeDate',
+                      'FileModifyDate', 'FileName', 'FilePermissions',
+                      'FileSize', 'FileType', 'FileTypeExtension',
+                      'MIMEType', 'NumChannels', 'SampleRate', 'SourceFile',
+                     }
+
+
+class AIFFParser(video.AbstractFFmpegParser):
+    mimetypes = {'audio/aiff', 'audio/x-aiff'}
+    meta_allowlist = {'AvgBytesPerSec', 'BitsPerSample', 'Directory',
+                      'Duration', 'Encoding', 'ExifToolVersion',
+                      'FileAccessDate', 'FileInodeChangeDate',
+                      'FileModifyDate', 'FileName', 'FilePermissions',
+                      'FileSize', 'FileType', 'FileTypeExtension',
+                      'MIMEType', 'NumChannels', 'SampleRate', 'SourceFile',
+                      'NumSampleFrames', 'SampleSize',
+                     }

libmat2/bubblewrap.py

+"""
+Wrapper around a subset of the subprocess module,
+that uses bwrap (bubblewrap) when it is available.
+
+Instead of importing subprocess, other modules should use this as follows:
+
+  from . import subprocess
+"""
+
+import os
+import shutil
+import subprocess
+import tempfile
+import functools
+from typing import Optional, List
+
+
+__all__ = ['PIPE', 'run', 'CalledProcessError']
+PIPE = subprocess.PIPE
+CalledProcessError = subprocess.CalledProcessError
+
+# pylint: disable=subprocess-run-check
+
+
+@functools.lru_cache(maxsize=None)
+def _get_bwrap_path() -> str:
+    which_path = shutil.which('bwrap')
+    if which_path:
+        return which_path
+
+    raise RuntimeError("Unable to find bwrap")  # pragma: no cover
+
+
+def _get_bwrap_args(tempdir: str,
+                    input_filename: str,
+                    output_filename: Optional[str] = None) -> List[str]:
+    ro_bind_args = []
+    cwd = os.getcwd()
+
+    # XXX: use --ro-bind-try once all supported platforms
+    # have a bubblewrap recent enough to support it.
+    ro_bind_dirs = ['/usr', '/lib', '/lib64', '/bin', '/sbin', '/etc/alternatives', cwd]
+    for bind_dir in ro_bind_dirs:
+        if os.path.isdir(bind_dir):  # pragma: no cover
+            ro_bind_args.extend(['--ro-bind', bind_dir, bind_dir])
+
+    ro_bind_files = ['/etc/ld.so.cache']
+    for bind_file in ro_bind_files:
+        if os.path.isfile(bind_file):  # pragma: no cover
+            ro_bind_args.extend(['--ro-bind', bind_file, bind_file])
+
+    args = ro_bind_args + \
+        ['--dev', '/dev',
+         '--proc', '/proc',
+         '--chdir', cwd,
+         '--unshare-user-try',
+         '--unshare-ipc',
+         '--unshare-pid',
+         '--unshare-net',
+         '--unshare-uts',
+         '--unshare-cgroup-try',
+         '--new-session',
+         '--cap-drop', 'all',
+         # XXX: enable --die-with-parent once all supported platforms have
+         # a bubblewrap recent enough to support it.
+         # '--die-with-parent',
+        ]
+
+    if output_filename:
+        # Mount an empty temporary directory where the sandboxed
+        # process will create its output file
+        output_dirname = os.path.dirname(os.path.abspath(output_filename))
+        args.extend(['--bind', tempdir, output_dirname])
+
+    absolute_input_filename = os.path.abspath(input_filename)
+    args.extend(['--ro-bind', absolute_input_filename, absolute_input_filename])
+
+    return args
+
+
+def run(args: List[str],
+        input_filename: str,
+        output_filename: Optional[str] = None,
+        **kwargs) -> subprocess.CompletedProcess:
+    """Wrapper around `subprocess.run`, that uses bwrap (bubblewrap) if it
+    is available.
+
+    Extra supported keyword arguments:
+
+     - `input_filename`, made available read-only in the sandbox
+     - `output_filename`, where the file created by the sandboxed process
+       is copied upon successful completion; an empty temporary directory
+       is made visible as the parent directory of this file in the sandbox.
+       Optional: one valid use case is to invoke an external process
+       to inspect metadata present in a file.
+    """
+    try:
+        bwrap_path = _get_bwrap_path()
+    except RuntimeError:  # pragma: no cover
+        # bubblewrap is not installed ⇒ short-circuit
+        return subprocess.run(args, **kwargs)
+
+    with tempfile.TemporaryDirectory() as tempdir:
+        prefix_args = [bwrap_path] + \
+            _get_bwrap_args(input_filename=input_filename,
+                            output_filename=output_filename,
+                            tempdir=tempdir)
+        completed_process = subprocess.run(prefix_args + args, **kwargs)
+        if output_filename and completed_process.returncode == 0:
+            shutil.copy(os.path.join(tempdir, os.path.basename(output_filename)),
+                        output_filename)
+
+        return completed_process

libmat2/epub.py

+import logging
+import re
+import uuid
+import zipfile
+import xml.etree.ElementTree as ET  # type: ignore
+from typing import Any, Dict
+
+from . import archive, office
+
+
+class EPUBParser(archive.ZipParser):
+    mimetypes = {'application/epub+zip', }
+    metadata_namespace = '{http://purl.org/dc/elements/1.1/}'
+
+    def __init__(self, filename):
+        super().__init__(filename)
+        self.files_to_keep = set(map(re.compile, {  # type: ignore
+            'META-INF/container.xml',
+            'mimetype',
+            'OEBPS/content.opf',
+            'content.opf',
+            'hmh.opf',
+            'OPS/.+.xml'
+            }))
+        self.files_to_omit = set(map(re.compile, {  # type: ignore
+            'iTunesMetadata.plist',
+            'META-INF/calibre_bookmarks.txt',
+            'OEBPS/package.opf',
+             }))
+        self.uniqid = uuid.uuid4()
+
+    def is_archive_valid(self):
+        super().is_archive_valid()
+        with zipfile.ZipFile(self.filename) as zin:
+            for item in self._get_all_members(zin):
+                member_name = self._get_member_name(item)
+                if member_name.endswith('META-INF/encryption.xml'):
+                    raise ValueError('the file contains encrypted fonts')
+
+    def _specific_get_meta(self, full_path, file_path) -> Dict[str, Any]:
+        if not file_path.endswith('.opf'):
+            return {}
+
+        with open(full_path, encoding='utf-8') as f:
+            try:
+                results = re.findall(r"<((?:meta|dc|cp).+?)[^>]*>(.+)</\1>",
+                                     f.read(), re.I|re.M)
+                return {k:v for (k, v) in results}
+            except (TypeError, UnicodeDecodeError):
+                return {file_path: 'harmful content', }
+
+    def _specific_cleanup(self, full_path: str) -> bool:
+        if full_path.endswith('hmh.opf') or full_path.endswith('content.opf'):
+            return self.__handle_contentopf(full_path)
+        elif full_path.endswith('OEBPS/toc.ncx'):
+            return self.__handle_tocncx(full_path)
+        elif re.search('/OPS/[^/]+.xml$', full_path):
+            return self.__handle_ops_xml(full_path)
+        return True
+
+    def __handle_ops_xml(self, full_path: str) -> bool:
+        try:
+            tree, namespace = office._parse_xml(full_path)
+        except ET.ParseError:  # pragma: nocover
+            logging.error("Unable to parse %s in %s.", full_path, self.filename)
+            return False
+
+        for item in tree.iterfind('.//', namespace):  # pragma: nocover
+            if item.tag.strip().lower().endswith('head'):
+                item.clear()
+                break
+        tree.write(full_path, xml_declaration=True, encoding='utf-8',
+                   short_empty_elements=False)
+        return True
+
+    def __handle_tocncx(self, full_path: str) -> bool:
+        try:
+            tree, namespace = office._parse_xml(full_path)
+        except ET.ParseError:  # pragma: nocover
+            logging.error("Unable to parse %s in %s.", full_path, self.filename)
+            return False
+
+        for item in tree.iterfind('.//', namespace):  # pragma: nocover
+            if item.tag.strip().lower().endswith('head'):
+                item.clear()
+                ET.SubElement(item, 'meta', attrib={'name': '', 'content': ''})
+                break
+        tree.write(full_path, xml_declaration=True, encoding='utf-8',
+                   short_empty_elements=False)
+        return True
+
+    def __handle_contentopf(self, full_path: str) -> bool:
+        try:
+            tree, namespace = office._parse_xml(full_path)
+        except ET.ParseError:
+            logging.error("Unable to parse %s in %s.", full_path, self.filename)
+            return False
+
+        for item in tree.iterfind('.//', namespace):  # pragma: nocover
+            if item.tag.strip().lower().endswith('metadata'):
+                item.clear()
+
+                # item with mandatory content
+                uniqid = ET.Element(self.metadata_namespace + 'identifier')
+                uniqid.text = str(self.uniqid)
+                uniqid.set('id', 'id')
+                item.append(uniqid)
+
+                # items without mandatory content
+                for name in ['language', 'title']:
+                    uniqid = ET.Element(self.metadata_namespace + name)
+                    item.append(uniqid)
+                break  # there is only a single <metadata> block
+        tree.write(full_path, xml_declaration=True, encoding='utf-8')
+        return True

libmat2/exiftool.py

+import functools
+import json
+import logging
+import os
+import shutil
+import subprocess
+from typing import Union, Set, Dict
+
+from . import abstract
+from . import bubblewrap
+
+
+class ExiftoolParser(abstract.AbstractParser):
+    """ Exiftool is often the easiest way to get all the metadata
+    from a import file, hence why several parsers are re-using its `get_meta`
+    method.
+    """
+    meta_allowlist: Set[str] = set()
+
+    def get_meta(self) -> Dict[str, Union[str, Dict]]:
+        try:
+            if self.sandbox:
+                out = bubblewrap.run([_get_exiftool_path(), '-json',
+                                      self.filename],
+                                     input_filename=self.filename,
+                                     check=True, stdout=subprocess.PIPE).stdout
+            else:
+                out = subprocess.run([_get_exiftool_path(), '-json',
+                                      self.filename],
+                                     check=True, stdout=subprocess.PIPE).stdout
+        except subprocess.CalledProcessError:  # pragma: no cover
+            raise ValueError
+        meta = json.loads(out.decode('utf-8'))[0]
+        for key in self.meta_allowlist:
+            meta.pop(key, None)
+        return meta
+
+    def _lightweight_cleanup(self) -> bool:
+        if os.path.exists(self.output_filename):
+            try:  # exiftool can't force output to existing files
+                os.remove(self.output_filename)
+            except OSError as e:  # pragma: no cover
+                logging.error("The output file %s is already existing and \
+                               can't be overwritten: %s.", self.filename, e)
+                return False
+
+        # Note: '-All=' must be followed by a known exiftool option.
+        # Also, '-CommonIFD0' is needed for .tiff files
+        cmd = [_get_exiftool_path(),
+               '-all=',         # remove metadata
+               '-adobe=',       # remove adobe-specific metadata
+               '-exif:all=',    # remove all exif metadata
+               '-Time:All=',    # remove all timestamps
+               '-quiet',        # don't show useless logs
+               '-CommonIFD0=',  # remove IFD0 metadata
+               '-o', self.output_filename,
+               self.filename]
+        try:
+            if self.sandbox:
+                bubblewrap.run(cmd, check=True,
+                               input_filename=self.filename,
+                               output_filename=self.output_filename)
+            else:
+                subprocess.run(cmd, check=True)
+        except subprocess.CalledProcessError as e:  # pragma: no cover
+            logging.error("Something went wrong during the processing of %s: %s", self.filename, e)
+            return False
+        return True
+
+@functools.lru_cache(maxsize=None)
+def _get_exiftool_path() -> str:  # pragma: no cover
+    which_path = shutil.which('exiftool')
+    if which_path:
+        return which_path
+
+    # Exiftool on Arch Linux has a weird path
+    if os.access('/usr/bin/vendor_perl/exiftool', os.X_OK):
+        return '/usr/bin/vendor_perl/exiftool'
+
+    raise RuntimeError("Unable to find exiftool")

libmat2/harmless.py

 import shutil
-from typing import Dict
+from typing import Union, Dict
 from . import abstract
 
 
 class HarmlessParser(abstract.AbstractParser):
     """ This is the parser for filetypes that can not contain metadata. """
-    mimetypes = {'text/plain', 'image/x-ms-bmp'}
+    mimetypes = {'text/plain', 'image/x-ms-bmp', 'image/bmp'}
 
-    def get_meta(self) -> Dict[str, str]:
+    def get_meta(self) -> Dict[str, Union[str, Dict]]:
         return dict()
 
     def remove_all(self) -> bool: