> No, treating some letter sequences like fi and ffi so that the letters stick together in a certain way is not "by this measure". Except ligatures are not lim

> No, treating some letter sequences like fi and ffi so that the letters stick together in a certain way is not "by this measure". Except ligatures are not limited to fi and ffi. They can be used to change appearance of any sequence of characters, including words. > It's noticeable; a B consistently occurs everywhere there is an A, wherever that font is used. I'm not talking about substituting a single letter. I'm talking about outright obfuscation of a entire texts. > You can't perpetrate an easter egg whereby a certain A is replaced. Quite the opposite, it is trivial to apply different fonts to different parts of text, so you can keep some parts of text normal, and some parts (like numbers and data) obfuscated. In HTML you may notice some abundance of <span>s, sure, but in PDF? Literally no way, unless you copy-paste everything and compare it. And this is not a hypothetical scenario or a 'minor threat'. Russian government used precisely this technique to obfuscate data in their online voters turn-out reports: (source in Russian, not found it reported in English) https://habr.com/ru/news/578846/ https://habr.com/ru/news/578846/. Have they used more subtle scrambling of only digits and not also mixing letters: who knows how long it would take for anybody to notice. From a technical perspective it's not so different from contextual alternates - both require custom made fonts. There are other ways to alter document presentation too. Especially in CSS, you can create pseudo-elements, not present in the original HTML, or hide parts of HTML, or use a media query which checks if the page is being printed or not! > We could prepare a page where every glyph appears, and detect substitutions via OCR. Why the extra step? Why not take the OCR of potentially compromised document and compare it with its raw content? It would detect every substitution, whether it's ligatures, alternates, media queries etc. Why not inspect the font file for defined alternates directly?