OCR giving out Random Characters

[Tsamsiyu1]

I have a couple documents (PDF's) that already have readable text. When i put them into paperless (it uses OCRmyPDF) and the OCR finishes it is not readable anymore.
This is a screenshot of part of the Original pdf.

and this is the screenshot of the same part but of the archive file:

my OCR is set to german and english and worked fine till now. all the other documents are processed correctly. just those 4 files look like this.
Is this a setting I have wrong or is the file itself protected?

[ke-ma-fi]

I am encountering an issue with invoices from a specific firm. While processing 4,000 documents in my system, I didn’t experience this problem with any other documents. However, invoices from this particular firm consistently display the same random characters, as previously described.

[bernd3k]

This has the feeling of a bad atttempt at an obfuscation layer à la ROT13 within the PDF - though it's not "shift by 13" in this case. Every symbol is replaced by the same other symbol in the entire document.

original char	ASCII code	replacement char	ASCII code	shift
A	65	$	39	26
b	98	E	69	29
w	119	Z	90	29
a	97	D	68	29
s	115	V	86	29
s	115	V	86	29
e	101	H	72	29
r	114	U	85	29

The A is shifted less because they might need the logical replacement "!" for something else. All numbers must have been shifted into the realm of control characters (unprintable) or filtered out entirely.

[jbarlow83]

When PDFs do not have a Unicode mapping table, each PDF reader has its own strategy to guess the Unicode mapping. Some guess poorly.

Also, some Unicode mapping tables are corrupt or incorrect, and then again, each PDF reader has its own strategy.