If you are using Proprietary Tools, you have no guarantee what the tools will be doing to "optimize" the workflow & what Meta Data gets retained in conversions.
Eg 1 : If there are layers of text below the layers of Images (to "assist" text-2-speech users) then whiteout may be undone.
Eg 2 : If the tools want to include text to enable "text search" of the Images , then the text may be there in some Meta Data like Comments or Annotations.
Eg 3 : Certain tools store revision history (to help "undo & to "audit") & these may leak the unwanted text.
Eg 4 : Some tools generate Caches & Indexes (to help users with quick output) which may reveal unwanted text.
The Best way (with a guarantee that the content is "gone") would be :
(1) Add the white-out (either by Placing Square on top layer or by Blackening the text)
(2) Convert the whiteout Pages to Images using some tools (Paranoid users might want to (2A) Check the generated Images & (2B) Eliminate all unwanted Meta Data , especially Comments & Annotations)
(3) Stitching these Images into a set of Pages to share & view.
UPDATE (to include Supporting Material) :
Since somebody thought that there is something wrong in this Answer.
The Almost Same Process is listed in this Document Page 11 :
...
the process of redacting in PDF involves:
•covering [[ Whiteout ]] each item of confidential information
with a black rectangle or by using black text highlighting
•converting [[ Image generation ]] the PDF document to multiple TIFF image files
•converting and reassembling [[ Stitching ]] the files into a single PDF document.
...
NOTE: Converting to TIFF and back to PDF has the unfortunate consequence that
the file will no longer be searchable, and accessibility is lost
because document structure and tags are lost in the process.
Using a third-party redaction plug-in would avoid such problems.
...
IMPORTANT: At this point, all you have done is to
cover up the confidential information.
To remove the information,
you need to “flatten” the file by converting to TIFF images
...
I have added some text & high-lighting :
The 3 Stages match the Process I outlined earlier.
The Issues mentioned are also what I listed earlier , that is searchability & accessibility getting lost.
While it says to use third-Party tools to avoid such Problems, I think using such third-Party tools would mean including the Sensitive Content in the Meta Data where some user can & will look to Extract. I think these should be avoided & "Images-only" will the safest way even with loss of searchability & accessibility.
The Pitfalls of "assuming redaction through software" are listed here :
(1) It lists Items related to (1A) Hillary Clinton / Sidney Blumenthal/ Libya , (1B) Apple / U.S. District Court , (1C) Citigroup / Social Security numbers & (1D) Paul Manafort / Russia.
(2) It also says these "obvious" methods are not effective or foolproof :
(2A) Changing the text's color to white. This may make it look as though the selected words to be redacted are hidden, but the remaining metadata can reveal the hidden text.
(2B) Blacking out with comment tools: Edits made by such tools can be removed to reveal the underlying text.
(2C) Deleting words or sections: Metadata contains document revision history and can be used to view deleted information.
(2D) Using dark tape or opaque marker: Rather than physically clipping out sensitive information, it is common practice to cover such information with dark tape or a marker and scan it into a PDF format. However, many scanners are sensitive enough to view such covered words even if they do not appear to be visible.
Putting all these together, I think the Process listed in this Answer is quite effective & foolproof !
