Remove pdf metadata

In order to remove metadata from PDF documents we will use the
exiftool_vendored gem.

The following line:
  Exiftool.new(attachment_path, "-overwrite_original -all:all=")
Overwrites the original file with another file without metadata.

So far this is the best solution we have found to perform this
metadata deletion.

When using Exiftool an exception is thrown, so we added a rescue
to handle it. Here is a task created where this problem is
discussed: exiftool-rb/exiftool.rb#28.
We'll wait to see if this will be fixed in future versions.

Translated with www.DeepL.com/Translator (free version)

Gemfile.lock

@@ -197,6 +197,10 @@ GEM
       multi_json (>= 1.3)
       rake
     execjs (2.8.1)
+    exiftool (1.2.4)
+      json
+    exiftool_vendored (12.60.0)
+      exiftool (>= 0.7.0)
     factory_bot (6.2.0)
       activesupport (>= 5.0.0)
     factory_bot_rails (6.2.0)
@@ -694,6 +698,7 @@ DEPENDENCIES
   devise-security (~> 0.16.0)
   email_spec (~> 2.2.0)
   erb_lint (~> 0.3.1)
+  exiftool_vendored (~> 12.60.0)
   factory_bot_rails (~> 6.2.0)
   faker (~> 2.22.0)
   file_validators (~> 3.0.0)

Gemfile_custom

@@ -24,4 +24,5 @@
 #
 # Add your custom gem dependencies here
 
+gem "exiftool_vendored", "~> 12.60.0"
 gem "ruby-ntlm", "~> 0.0.4"

app/models/document.rb

@@ -11,6 +11,8 @@ class Document < ApplicationRecord
 
   scope :admin, -> { where(admin: true) }
 
+  before_save :remove_metadata
+
   def self.humanized_accepted_content_types
     Setting.accepted_content_types_for("documents").join(", ")
   end
@@ -36,4 +38,14 @@ def association_name
     def documentable_class
       association_class
     end
+
+    def remove_metadata
+      return unless self.attachment.attached?
+      attachment_path = ActiveStorage::Blob.service.path_for(self.attachment.key)
+
+      Exiftool.new(attachment_path, "-overwrite_original -all:all=")
+      self.attachment.attach(io: File.open(attachment_path), filename: self.attachment.filename.to_s)
+    rescue Exiftool::ExiftoolNotInstalled
+      nil
+    end
 end