- From: Jonathan Kew <jfkthame@gmail.com>
- Date: Mon, 14 Jan 2019 13:33:08 +0000
- To: public-webfonts-wg@w3.org
On 14/01/2019 12:44, Levantovsky, Vladimir wrote: > A: When a major security research company discovers that scumbags are > now using it for their phishing campaigns! > (https://www.siliconrepublic.com/enterprise/phishing-web-fonts-fake) This doesn't seem particularly interesting to me. It sounds like they used a (web)font with a scrambled encoding in order to disguise content that otherwise might have been noticed by some kind of scanner or security product; but there are already countless ways of obfuscating the content that a user will end up seeing, so that scanning page sources isn't much of a defence anyway. E.g. if I want to create a phishing page for Big Bank Corp, but I don't want a security product to see this string when it scans the source of the page, I could just use Unicode to do ‮giB‬‎ ‮knaB‬‎ ‮proC‬ instead, without needing JS hackery or even CSS to disguise the text. Or for a CSS hack that doesn't require a webfont to disguise the text, we could do <style> .a::before { content: "B" } .a::after { content: "g" } .b::before { content: "B" } .b::after { content: "k" } .c::before { content: "C" } .c::after { content: "p" } </style> <span class=a>i</span> <span class=b>an</span> <span class=c>or</span> Good luck detecting that with a page-source scanner. So - sure, a custom webfont can be used to "disguise" text so that its apparent content when seen by the human visitor is quite different from the underlying encoded text. That's hardly news - we were abusing custom fonts like that years before Unicode was even part of the game - and doesn't significantly change the security landscape, afaics. If someone were able to use a webfont to change the display of URLs (or page titles, etc) within the browser UI (rather than the content of the page), that would be a different story - although if that were to happen, I'd regard it as a browser bug rather than a flaw in the webfont technology. JK > > I must admit that using webfonts as a substitution cypher is a clever > idea, and I can see some potentially good uses for it (imagine building > a secure communication channel where a cypher is switched > algorithmically by e.g. using different font style/weights), but it also > begs another question to be asked – were we too optimistic when we > declared DSIG to be of no significant importance for webfonts / WOFF2? Suppose we required webfonts to have valid DSIG signatures, or something like that. What difference would it make here? None, afaict; the bad guy could just sign the font and proceed in exactly the same way. A signature does nothing to guarantee that the font resource isn't going to mislead the reader. > And, do we need to update “Security considerations” section knowing that > webfonts could be a much more treacherous grounds than we previously > imagined? > > Thanks, > > Vlad >
Received on Monday, 14 January 2019 13:33:33 UTC