[dunham]

Hmm, I guess I've been using that term incorrectly to describe fixed substitution ciphers. It's been a few decades since I learned this stuff.

It's effective at preventing a simple copy/paste or "view source" but ineffective against an adversary of any sophistication.

I've seen this used in PDF files before. There, if you have a pristine copy of the embedded font, you can just look up a copy of the character program to find the original value. But it's effective against naive pdf to text or pdf to html services.

Then again, you could just ocr the rendered text, so I guess the obfuscation technique only has to be as effective as rendering the text as a bitmap.

[bentcorner]

Yeah, exactly. I OCRd the first line of the text and manually decrypted the first line of the book just to verify that's what was happening. You could mitigate it by having more complex font files with multicharacter glyphs and include non printing characters in the stream. Probably a "good enough" solution to stop the copy/pasting folks. Anybody knowledgable enough to crack this is is probably not going to waste the time and torrent the book anyway.