OT: Art of Electronics, 3rd ed, for $59.96

Thank-you for your uninformed opinion; we're all dumber for it. You might like to Google "steganography".

Speaking as someone who has written a PDF parser, and spent a decade building banking-grade cryptographic products... the "easy" methods for watermarking are easy to remove, but it's actually not hard to put either visible or invisible watermarks into a PDF so that they're really hard to remove.

I won't go into all the details, but there are at least a thousand ways to encode data into PDF without changing the rendered result. If random variations are included in many places, then simple differencing cannot find where the watermark is stored. And if the watermark is extended across all these locations using heavily redundant FEC, then the watermark is effectively impossible to remove.

Has this actually been done? I'm not sure... but I'm sure it can be. Yes, it's harder than appending an object to the PDF dictionary or inserting text into the MSWord original, but not much harder.

How is watermarking annoying? Besides, the main benefit of watermarking isn't that it enables prosecution, but that it discourages sharing in the first place. The buyer doesn't need to see the watermark on every other page, they just need to know that their purchase can be traced.

I'm not talking about enforced DRM. I personally agree with you here.

Clifford Heath.

IME it's pretty straightforward to remove the watermark from, for example, papers that have the download IP embedded.

Certainly the people who throw this stuff onto the net in bulk would hardly be slowed down.

--sp

Yup, no dice for us ordinary folks :-(

Hey, they even have a photo of Winfield with a much shorter beard than on the 2nd edition dust cover:

Trivial watermarking is trivially removed, duh. But I wasn't suggesting that, read my other reply.

OK here's my picture of AoEX online. You ask everyone to buy in, a one time payment. ~$100 seems to me about right. (You could have student members for ~$10) Then you also make it totally open with ~30-90?? day guest memberships. Right scofflaws can abuse you... Perhaps you can shame some of them into paying? (I give money to Wikipedia.) George H.

They're thinking that anyone who buys the paper book gets a free live html book. They're also thinking of bundling AoE 3 with the x-Chapters. You'd have one big live well-linked document. The paper versions would gradually become out of date.

Of course AoE2 was right up to the minute in 2014. ;)

Cheers

Phil "hanging on to all 3 editions" Hobbs

Of course AoE2 was right up to the minute in 2014. ;)

Cheers

Phil "hanging on to all 3 editions" Hobbs

AoE 3 you mean? It was up-to-date a year before it was published.

In the first edition he looked like Severn Darden (GREAT actor).

No, AoE2. I wasn't being snippy, I was just teasing you about the long wait . ;)

Cheers

Phil

I am aware of steganography.

I think that visible watermarks may be difficult to remove if they are overlapping the content that the purchaser paid to see (but may be either annoying to paying customers, or otherwise non-trivial to generate for arbitrary content).

Truly invisible watermarks should be very easily removed by rendering the page to a bitmap, quantizing the pixel colours (usually black and white would do), and then optionally re-encoding that to a PDF if you still want it in pdf file format. I would not call that "really hard". If you want the text to be searchable then there is the extra step of running it through OCR, which may be difficult to do well, but I doubt many pirates would be that bothered by the accuracy of the OCR text. Some documents are so poorly structured (e.g. text stored as bitmap images in the original document, or parts of words in separare strings where someone was trying to do custom character spacing) that the rendering+OCR process might sometimes improve the usability!

True. Of course these are all useless if the pirate uses only the rendered (bitmap) result.

Yes, this could work, but is not really trivial to implement. If the document is valuable enough to warrant this kind of work (e.g. drafts of a free trade agreement during the phase after the pharma companies and MPAA finished writing it, but before it has been signed and shown to the citizens whose rights it is about to delete...), then the person pirating/leaking it might also go to some trouble. For example, if the pirate has multiple copies of the document with different watermarks and randomly uses pixels from different copies for different parts of each page, the FEC might need to be quite good to withstand that.

Personally I have never seen any watermarking that affects the pixel colours in the rendered result, other than very obvious text on the edge of each page.

Not often, but I have had downloads take a long time to start, and I think this was due to the machine that does the watermarking being not up to the load.

It could also be annoying if your coworker (with whom you are legally permitted to share a document) then gives it to someone else. In other words, the person whose name is on the watermark is not necessarily the one who did any copyright infringement. To avoid this, you could make your coworkers each download their own copy, rather than using shared drives or dropbox etc. (or even shared paper copies) for sharing documents for which you have a company-wide licence, but that would be annoying. Some watermarks are vague enough that the individual user within the company cannot be identified, however I suspect that this would destroy most of their deterrent effectiveness.

Ok, I should not have conflated DRM with watermarking.

Win made it past 65 though and seems to be in good health.

There are steganographic methods that survive converting to 1-bit monochrome, cropping, and stretching. The ones I know about encode the message in the low spatial frequencies of an image. The work was done by some IBM colleagues of mine around 2000.

Cheers

Phil Hobbs

In 1999 I saw a watermark added to a professional quality jpeg photo. Then the image was shrunk and the quality reduced, then printed on an inkjet printer and scanned.

The watermark survived.

But watermarking etc is pretty much irrelevant unless there are effective sanctions against copying and/or distributing. In most cases, there aren't.

It is another variant of the old axiom "if you think cryptography will solve your problem, then you don't understand cryptography and you don't understand your problem".

I have heard that color copy machines add encoded IDs, so the feds can tell which copier made those counterfeit $100 bills.

Your inkjet printer probably puts faint yellow dots in white areas that encode the same info. Not all do, but afaik most.

Cheers

Phil Hobbs

Bit old and written by a Physicist. Many better books nowadays

Ahem! If you're referring to the subject book, AoE, 3rd ed, it's not very old, ink barely dry, and virtually error-free copies only available for about a year. I'm a circuit designer electronics engineer, not a physicist.