Watermark for EPUB - Page 2

Chang · 03-25-2010, 03:25 AM

Quote:

Originally Posted by larkki

Have you proceeded with your solution? I'm interested because I was looking for some solutions from the web and so far haven't been able to find any.
What are the combinations that you are using and how's it working so far?

Yes, I have proceeded with my solution but I don't have any experience on how it will work in practice, yet. I won't tell my exact ways to do the watermark but I mostly used the tips in this thread and some few other ideas as well. I don't know how my solution will work but I guess I'll find it out in the next couple months.

darkmonk · 03-29-2010, 12:16 AM

I have a devilish solution, but it requires some programming. Dynamical generate each epub. I;ve though of several places you can add your information: The names of the directories and files, the names of the attributes (class, div, etc.), or, more evil-y, edit the fonts. Add a series of characters in there that is always the email or something. As for obscuring the data you want to add, which you should do to make it non-obvious, simply encrypt each piece. Oh, another thing I've heard of it directly adding on a zip file to a picture. It's supposed to be ignored, and then by recording the byte size you can just take that much of the end to recover it.

That'd take some work, but each should work fine. Positively brilliant, methinks. But I doubt you will implement any of those methods. Another would be using svg images with text embeded in them set to display at 100% transparency; svg's can even have fonts, so you could hide one there. The possibilities are endless!

Fat Abe · 03-29-2010, 06:29 AM

The jpeg route allows one to insert a number of telltales, and that is the most inconspicuous, unless the thief had two copies of the book. That still seems the most preferable for an average person.

I'm not that familiar with the spec on the opf file, but can't one slip in a block like <xyz><item id="abcd1234" /></xyz>, where xyz can be any string other than metadata, manifest and the other reserved names? Or just obfuscate by using a known name type where assignment of the id does not conflict with anything else. id is used to hold the hex value of the registered user.

As we come up with new identity embedding methods, one can be sure the pirates are working on stripping programs to automatically eradicate all ownership tags, including the ISBN.

larkki · 03-29-2010, 06:54 AM

Good ideas people! I was thinking about something along these lines as well.
But do you know any commercial or open source solution which would provide epub watermarking? Or companies who are working on such technology? I'm doing my master's thesis on the subject so that's why I'm curious.

DaleDe · 03-29-2010, 11:17 AM

Quote:

Originally Posted by larkki

Good ideas people! I was thinking about something along these lines as well.
But do you know any commercial or open source solution which would provide epub watermarking? Or companies who are working on such technology? I'm doing my master's thesis on the subject so that's why I'm curious.

One of the problems is that if you tell everyone how you did it then it compromises the security of the method. This is why there is no public domain DRM schemes. Many of the proposed solutions in this thread were open source but they depend on the end user not knowing what the originator did. Obscurity is essential unless you are just putting a real watermark in the document like a background image on every page or including the users name at the top of every page.

Dale

darkmonk · 03-29-2010, 08:41 PM

Yes - I want to say, there is no method you could ever do that cannot be removed. The most difficult method would simply be a direct alteration of the zip file, that did not impact any of the source files. It would be an incredible exploit, but good luck finding it.

Fat Abe · 03-29-2010, 10:39 PM

Quote:

Originally Posted by DaleDe

One of the problems is that if you tell everyone how you did it then it compromises the security of the method.

Dale

Most of the ideas mentioned in this thread have been tried out already. The second layer of defense in any DRM'ed document are these telltales, which exist in jpeg, opf, html, xml, and css files. I have a theory about why Adobe and Amazon made it so easy for their DRM schemes to be broken. It's because they stuck enough telltales into the files to trace it back to the source recipient, who was then dumb enough to post it on the internet. OK, maybe I'm giving Adobe too much credit. It could be that they really are negligent. The next revisions of ADE and Amazon's Kindle for PC should be very interesting.

In the html files, I'm waiting to see if an author tries adjusting the words between the <p> and </p> paragraph tags, using ordinary line feeds to vary the count of words per line in the ascii file. And to do it "uniquely" per delivered file. Such a change would be totally transparent to any ereader, and might even look natural to someone editing the raw html. Of course, this scheme is easily defeatable.

The OP has been exposed to a number of good ideas by this board, and let's hope he can get back to us in a few months to see if he caught a pirate distributor at work.

Jellby · 03-30-2010, 03:48 AM

Quote:

Originally Posted by darkmonk

Yes - I want to say, there is no method you could ever do that cannot be removed. The most difficult method would simply be a direct alteration of the zip file, that did not impact any of the source files. It would be an incredible exploit, but good luck finding it.

I think you can just append extra bytes to a valid zip file, with the zip file remaining valid and the extra bytes not being seen anywhere, unless you look for them with an hex editor or similar. Of course, if someone re-creates the zip (because they have modified the metadata to have "Surname, Firstname" instead of "Firstname Surname", for instance), the bytes are lost.

03-29-2010, 06:29 AM	#18
Fat Abe Man Who Stares at Books Posts: 1,816 Karma: 10606722 Join Date: Mar 2010 Location: 50th State, USA. Also, PA, NY, CA, and elsewhere. Device: All of the Above	The jpeg route allows one to insert a number of telltales, and that is the most inconspicuous, unless the thief had two copies of the book. That still seems the most preferable for an average person. I'm not that familiar with the spec on the opf file, but can't one slip in a block like <xyz><item id="abcd1234" /></xyz>, where xyz can be any string other than metadata, manifest and the other reserved names? Or just obfuscate by using a known name type where assignment of the id does not conflict with anything else. id is used to hold the hex value of the registered user. As we come up with new identity embedding methods, one can be sure the pirates are working on stripping programs to automatically eradicate all ownership tags, including the ISBN. Last edited by Fat Abe; 03-29-2010 at 06:39 AM. Reason: typo

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
epub, ePub, EPUB, warum blos ePub?	flowoeB	Lounge	5	11-27-2009 09:37 AM
Normal to see an Adobe logo "watermark" while charging?	rahulm	Sony Reader	3	03-16-2009 04:24 AM
Canon camera manual watermark	HappyMartin	iRex	2	10-30-2007 04:38 AM

03-29-2010, 12:16 AM	#17
darkmonk Connoisseur Posts: 58 Karma: 12 Join Date: Jan 2009 Device: none	I have a devilish solution, but it requires some programming. Dynamical generate each epub. I;ve though of several places you can add your information: The names of the directories and files, the names of the attributes (class, div, etc.), or, more evil-y, edit the fonts. Add a series of characters in there that is always the email or something. As for obscuring the data you want to add, which you should do to make it non-obvious, simply encrypt each piece. Oh, another thing I've heard of it directly adding on a zip file to a picture. It's supposed to be ignored, and then by recording the byte size you can just take that much of the end to recover it. That'd take some work, but each should work fine. Positively brilliant, methinks. But I doubt you will implement any of those methods. Another would be using svg images with text embeded in them set to display at 100% transparency; svg's can even have fonts, so you could hide one there. The possibilities are endless!

03-29-2010, 06:54 AM	#19
larkki Junior Member Posts: 2 Karma: 10 Join Date: Mar 2010 Device: iphone	Good ideas people! I was thinking about something along these lines as well. But do you know any commercial or open source solution which would provide epub watermarking? Or companies who are working on such technology? I'm doing my master's thesis on the subject so that's why I'm curious.

03-29-2010, 08:41 PM	#21
darkmonk Connoisseur Posts: 58 Karma: 12 Join Date: Jan 2009 Device: none	Yes - I want to say, there is no method you could ever do that cannot be removed. The most difficult method would simply be a direct alteration of the zip file, that did not impact any of the source files. It would be an incredible exploit, but good luck finding it.