Rogue characters corrupting AcroForm object

I've used PDFBox (2.0.7) to develop an interactive PDF document.

When completed, the user uploads the PDF to an app server.

In the many hundreds that have been successfully filled-out and uploaded, I have had three instances where a rogue character has corrupted the AcroForm object in the PDF causing the upload to fail.

In two cases the rogue character is in the structure of the AcroForm object, such that I cannot read data from that line of the object.

 

The other time, the rogue character was appended into a data value of the AcroForm object.

Although I could read the data value, it was not the value that I wrote out when creating the PDF file.

 

I have attached three screen-shots showing the rogue characters.

 

Has anyone else experienced the same issue?

I've looked through the web and not found anything similar.

 

I'm confident that the rogue characters were not there when the PDF was created.

It is though they are put there while the user interacts with the PDF.

The user may open and close the file multiple times before upload.

 

Cheers

Francis