Content Amazon Watermarking eBooks?

[JSWolf]

Originally Posted by Dyxan

Hello all,
I happen to purchase few ebooks in mobi and azw3 formats and than to convert them in calibre to PDF format.
In the PDF file to my surprise i found my name and email adress identification like (purchased at on ... by.. Name.. email adress). I do not like to have this identification there, is there any simple 1-2 click way to remove this identification allready in the AZW3 or mobi original file?
Sure I can remove it f.e. manually by INFIX in the resulting PDF, though I would like to do it allready in the source file...

thx in advance for any usefull ideas

Has Amazon started watermarking eBooks?

[jhowell]

Originally Posted by JSWolf

Has Amazon started watermarking eBooks?

Someone will need to provide more evidence before I will believe that a new type of readable watermark that survives conversion using calibre has been added by Amazon.

(Amazon has long added non-readable watermarks to the EXTH metadata in Kindle books.)

[JSWolf]

Originally Posted by jhowell

Someone will need to provide more evidence before I will believe that a new type of readable watermark that survives conversion using calibre has been added by Amazon.

(Amazon has long added non-readable watermarks to the EXTH metadata in Kindle books.)

But when you remove the DRM, that type of watermark is very easy to remove.

I've not see any visible watermark when I've read Kindle eBooks converted or shifted to ePub.

[barryem]

I just tried converting a Kindle book to PDF and searching the PDF for my name and email address, etc. I couldn't find it.

Barry

[DNSB]

The only time I've seen such a watermark was a free book my wife downloaded from Entangled. The pdf and the mobi converted to pdf or epub featured a graphic with her email address on every page. The nasty bit when I converted to epub was that every page graphics converted to a separate file which might have explained why the original mobi was so large.

I suspect that the books were likely not watermarked by Amazon but rather by the original publisher or that they were not purchased through Amazon -- other vendors do support mobi/azw3 direct purchases.

For those who are interested, the original thread is found at: how remove text from purchased mobi or azw3?.

[lumpynose]

Doitsu's answer is exactly what I was thinking. And does amazon still sell books in mobi format?

I also find using the word watermark confusing; to me a watermark is some semitransparent text (or image) overlaying the original text, like a big "Draft" going from corner to corner. The original post from dyxan didn't use the word watermark.

[haertig]

I just checked a book that I bought from Amazon about 3 months ago. There is no trace of a human-readable version of my credentials in either the DRM'ed azw3 file or in the Calibre DRM-stripped version of the file. Note: Both my Amazon credentials and my email address contain a unique text string that I searched for, so you'd expect that might be what shows up, if an identifier was there.

I'm not saying that an identifier is not there - it could be stored in some non-human-readable form (encrypted) - I'm just saying that it is not there in plain ASCII text form.

Test that was run from Linux:

"strings filename.azw3 | grep -i unique_string" (where "unique_string" is a string of letters found in my amazon credentials and also in my email address on file with Amazon)

I searched "filename.azw3" as it existed straight from Amazon, before import into Calibre, and also after Calibre had de-DRM'ed it and stored it in the Calibre file structure.

So if Amazon is indeed watermarking book purchases, they're being sneaky in the way they do it. Encrypting or in some other way scrambling the watermark. Or they could be using some other identifier - maybe some internal database row_id that uniquely identifies me in a way that only Amazon would know.

There is this one slightly suspicious looking text section found in the ebook azw3 file which could be an encrypted watermark:

Ariel
EBOK
atv:kin:2:ihS5FEWN2tLW9uSBqK26KD3Lhf8ZAE/kZQboZ9lrS3ZEIkMD0HOtwIFE/SNthee6rvhziZWx4TSxfWCftHj8+PUHAEqBMY5cdYM7wwBzlTYwNNTiydy1UEBKfQNbvqkGaK8yRdxjAqRDiUGhQ7Vx11kvhp+jw+K092Etcs21igA=:/Ty5lDTBtCXV78x859TIchV4ovw=
Ariel

Note that the book I bought was "Ariel", by Steven R. Boyett - that text shows up both before and after the suspicious "atv:kin:2" string. I didn't find anything else that aroused my suspicion as a watermark in the ebook. I guess if one were paranoid, I would try manually editing out the atv:kin:2 line and then test to see if the ebook still functions normally. The potential exists that you could really bork up an ebook by manually removing lines of text that you don't understand the structure of, so beware.

[lumpynose]

.....

[lumpynose]

Originally Posted by haertig

So if Amazon is indeed watermarking book purchases, they're being sneaky in the way they do it. Encrypting or in some other way scrambling the watermark. Or they could be using some other identifier - maybe some internal database row_id that uniquely identifies me in a way that only Amazon would know.

But his email and name appeared after he converted it to PDF with calibre; that implies that calibre knows how to get at this sneaky data, which I think is unlikely even if it is hidden away. Look at doitsu's answer, that seems like the logical conclusion, that he bought these books from somewhere other than amazon.

[stumped]

off topic slightly but amazon "watermark" the digital music files they sell ( with unique identifiers in MP3 tags). removable but not trivial as they use tags which most tag editors do not display.
so I am not at all suprised if they do something similar with books
Their digital video is encrypted and pretty unbreakable, except if you want to spend an hour re-recording each hour of video via grab playback software

[ i have not bought an mp3 from amazon for ages, there used to be a freeware amazon tag remover but it stopped working after some windows update, dunno if one exists now ]
FWIW I have never read of amazon taking a customer to court for uploading purchased digital content

[JSWolf]

Originally Posted by DNSB

The only time I've seen such a watermark was a free book my wife downloaded from Entangled. The pdf and the mobi converted to pdf or epub featured a graphic with her email address on every page. The nasty bit when I converted to epub was that every page graphics converted to a separate file which might have explained why the original mobi was so large.

I suspect that the books were likely not watermarked by Amazon but rather by the original publisher or that they were not purchased through Amazon -- other vendors do support mobi/azw3 direct purchases.

For those who are interested, the original thread is found at: how remove text from purchased mobi or azw3?.

I know other vendors/publishers besides Amazon still support Mobi. When Tor does a free giveaway, the Mobi that they give is a combined KF8/Mobi.

Who other then Tor gives and/or sells eBooks in KF8 outside of Amazon?

[JSWolf]

Originally Posted by lumpynose

Doitsu's answer is exactly what I was thinking. And does amazon still sell books in mobi format?

Mobi aka Mobipocket format is delivered to Kindles (Kindle 1, Kindle 2, Kindle DX/DX-G) that cannot handle KF8. This way, Amazon can keep supporting these obsolete devices with an obsolete format.

I also find using the word watermark confusing; to me a watermark is some semitransparent text (or image) overlaying the original text, like a big "Draft" going from corner to corner. The original post from dyxan didn't use the word watermark.

In this case, a watermark is used to embed something that can be used as an identifier. In this case, the watermark is used in an eBook to identify who purchased it. I know where Amazon stores the watermark they use (it's in the OPF) and it is very easy to remove if the file KF8 without DRM. But I have never seen or previously heard of Amazon using a watermark that is your name and email address.

[latepaul]

Originally Posted by haertig

Test that was run from Linux:

"strings filename.azw3 | grep -i unique_string" (where "unique_string" is a string of letters found in my amazon credentials and also in my email address on file with Amazon)

I'm not an expert on the AZW3 format but I think parts of it are compressed so 'strings' won't show you everything. Probably best to use Kindleunpack and grep the individual files.

[DNSB]

Originally Posted by haertig

There is this one slightly suspicious looking text section found in the ebook azw3 file which could be an encrypted watermark:

Ariel
EBOK
atv:kin:2:ihS5FEWN2tLW9uSBqK26KD3Lhf8ZAE/kZQboZ9lrS3ZEIkMD0HOtwIFE/SNthee6rvhziZWx4TSxfWCftHj8+PUHAEqBMY5cdYM7wwBzlTYwNNTiydy1UEBKfQNbvqkGaK8yRdxjAqRDiUGhQ7Vx11kvhp+jw+K092Etcs21igA=:/Ty5lDTBtCXV78x859TIchV4ovw=
Ariel

When I check the opf file from an unpacked AZW3, I find the following string:

<meta name="Watermark_(hex)" content="6174763a6b696e3a323a6c49706e36584466762b755036746f70392f57464d4f654a65625a51426c71357a7032504f58486b434b6b314c4a4d4b5871375a467856304a4972506c5463706d71506c59777a7464617046654b6846304e43716448505144562f3468726e55377342317754786e7173744173484b4870577772537a6b3874584d346167434f2b543035683776776a7930374e324a2f614c784f443733766b43614d386a4435593751326230396d533146473878725377743732553932354d53456b44594d793a384f62756f55575656476c4e4f5a4d7549345030746a63305777453d" />

which hex=>Ascii gives:

atv:kin:2:lIpn6XDfv+uP6top9/WFMOeJebZQBlq5zp2POXHkCKk1LJMKXq7ZFxV0JIrPlTcpmqPlYwztdapFeKhF0NCqdHPQDV/4hrnU7sB1wTxnqstAsHKHpWwrSzk8tXM4agCO+T05h7vwjy07N2J/aLxOD73vkCaM8jD5Y7Q2b09mS1FG8xrSwt72U925MSEkDYMy:8ObuoUWVVGlNOZMuI4P0tjc0WwE=

There was a discusion back in 2012 about the watermark entry and what was reported then was that the book ID and user ID were encoded in the string. From that discussion, I suspect the the string LJMKXq7ZFxV0JIrPlTcp which seems to be consistent in the 4 azw3 I've checked represents a user ID.

[lumpynose]

Originally Posted by DNSB

When I check the opf file from an unpacked AZW3, I find the following string:

So if someone wanted to remove that, then they could open the opf file in calibre's editor and delete it?

With calibre's editor I checked several books that I've bought from amazon and none of them have it. I was looking in the file metadata.opf. But then when I buy a book I go through some rigmarole because I'm obsessive about having ragged right margins. I download the AZW3 file from their web page, add it to my calibre library which removes the DRM, then convert it to EPUB, then remove the AZW3, then convert the EPUB to AZW3.

Other than removing the DRM I'm not doing anything sketchy with bought books so having my email embedded in the book doesn't bother me.

And I repeat my question about calibre's conversion to PDF. Does Dyxan's PDF show his email because calibre extracted it from that string? I don't find that easy to believe. I still think his information was in the AZW3 unencrypted and plain text and he didn't notice it, e.g., it was on page 1 or something (I'm thinking of the thread here about how the book cover isn't displayed when it's on page 1).