Как изменить md5 сумму файла

I need to duplicate various kinds of file types, change them a bit so that the original’s md5 hash won’t match the modified one, but keep them readable and not corrupted.

TXT files — that’s obvious. I just add a random string to the end of the file.

PDF file — well I started looking for a java library to edit pdf files, but then I accidentally tried to open a pdf file in notepad++, and thought — why don’t I try to add a random string to the end of the not readable content that I see there. Well, to my surprise it worked and the file wasn’t corrupted.

ZIP file — I’ve tried the same that I did with pdf, and it also worked.

DOCX- the same method stopped working here. Appending just a space (» «) at the end of the binary content of a docx file that I open in a text editor, corrupts the file.

So what I need is:

1. java libraries for modifying office documents :doc, docx, xls, xlsx, ppt, pptx.

2. There are still file types that I need to change there md5 hash output, but I don’t think they are modifiable in java — media files for example, executables and etc..
   So, nevertheless, how can i perform what I want on these files? Is there a way to just «touch» the file, change a header or something and make it nonidentical to an untouched one?

edit:
Ok, here’s the motivation — I want to generate massive amount of data as I asked here: How to produce massive amount of data?

At the time of that question, the answers I got there were enough, but not they dont.

1. I need the data to be nonidentical. Pairs of files must fail md5 hash test.

2. i can’t just generate random strings, because I need to simulate real files and documnets.

3. I can’t use existing data dumps, because I need various sizes of these data sets that include various file types. I need something that I’ll give as an input the size, and it will generate the data for me.

So I figured that I should use a starting data set of all the file types that I eventually need, and just duplicate this data set.

I need to duplicate various kinds of file types, change them a bit so that the original’s md5 hash won’t match the modified one, but keep them readable and not corrupted.

TXT files — that’s obvious. I just add a random string to the end of the file.

PDF file — well I started looking for a java library to edit pdf files, but then I accidentally tried to open a pdf file in notepad++, and thought — why don’t I try to add a random string to the end of the not readable content that I see there. Well, to my surprise it worked and the file wasn’t corrupted.

ZIP file — I’ve tried the same that I did with pdf, and it also worked.

DOCX- the same method stopped working here. Appending just a space (» «) at the end of the binary content of a docx file that I open in a text editor, corrupts the file.

So what I need is:

1. java libraries for modifying office documents :doc, docx, xls, xlsx, ppt, pptx.

2. There are still file types that I need to change there md5 hash output, but I don’t think they are modifiable in java — media files for example, executables and etc..
   So, nevertheless, how can i perform what I want on these files? Is there a way to just «touch» the file, change a header or something and make it nonidentical to an untouched one?

edit:
Ok, here’s the motivation — I want to generate massive amount of data as I asked here: How to produce massive amount of data?

At the time of that question, the answers I got there were enough, but not they dont.

1. I need the data to be nonidentical. Pairs of files must fail md5 hash test.

2. i can’t just generate random strings, because I need to simulate real files and documnets.

3. I can’t use existing data dumps, because I need various sizes of these data sets that include various file types. I need something that I’ll give as an input the size, and it will generate the data for me.

So I figured that I should use a starting data set of all the file types that I eventually need, and just duplicate this data set.