CSS tracking: Difference between revisions
m Flow of reading, tone moderation to neutrality (see Neutral POV / Editorial guidleines), etc. See Discussion page for more. |
mNo edit summary |
||
| (One intermediate revision by one other user not shown) | |||
| Line 6: | Line 6: | ||
CSS can declare that certain resources/assets be used ''if'' certain conditions are met.<ref>https://developer.mozilla.org/en-US/docs/Web/CSS/Guides/Media_queries</ref> Since browsers implement [[wikipedia:Lazy_loading|lazy-loading]], this means that assets will only be requested ''when'' the conditions are met. This effectively allows [[wikipedia:Ping_(networking_utility)|pinging]] arbitrary URLs when a client-side event happens. Instead of referencing a single endpoint for all events, each event can be associated to a different URL, allowing the tracking-"[https://www.gnu.org/philosophy/who-does-that-server-really-serve.html server]" to gather more data about user behavior.<ref>https://www.mike-gualtieri.com/posts/stealing-data-with-css-attack-and-defense</ref><ref>https://portswigger.net/research/inline-style-exfiltration</ref> | CSS can declare that certain resources/assets be used ''if'' certain conditions are met.<ref>https://developer.mozilla.org/en-US/docs/Web/CSS/Guides/Media_queries</ref> Since browsers implement [[wikipedia:Lazy_loading|lazy-loading]], this means that assets will only be requested ''when'' the conditions are met. This effectively allows [[wikipedia:Ping_(networking_utility)|pinging]] arbitrary URLs when a client-side event happens. Instead of referencing a single endpoint for all events, each event can be associated to a different URL, allowing the tracking-"[https://www.gnu.org/philosophy/who-does-that-server-really-serve.html server]" to gather more data about user behavior.<ref>https://www.mike-gualtieri.com/posts/stealing-data-with-css-attack-and-defense</ref><ref>https://portswigger.net/research/inline-style-exfiltration</ref> | ||
Traditionally, CSS tracking was (and still is) implemented as a limited [[Fingerprinting|finger printer]], typically by enumerating installed fonts and checking window dimensions.{{Citation needed}} | Traditionally, CSS tracking was (and still is) implemented as a limited [[Fingerprinting|finger-printer]], typically by enumerating installed fonts and checking window dimensions.{{Citation needed}} | ||
Either way, the attack has limitations, as [[wikipedia:Cache_(computing)|caching]] avoids (no guarantee) repeated requests from happening. | Either way, the attack has limitations, as [[wikipedia:Cache_(computing)|caching]] avoids (no guarantee) repeated requests from happening. | ||
==Why it is a problem== | ==Why it is a problem== | ||
Though CSS is widely believed to be "just a declarative styling | Though CSS is widely believed to be "just a declarative styling system" with no practical compute power, it is actually a [https://gavinhoward.com/2024/03/what-computers-cannot-do-the-consequences-of-turing-completeness/#mathematical-vs-practical virtually Turing-complete] programming language,<ref>https://lyra.horse/x86css/</ref><ref>https://lyra.horse/css-clicker/</ref> a fact which may leave even the most privacy-minded users vulnerable to tracking.{{Citation needed}} This mode of attack breaks the common belief that [[wikipedia:HTML|HTML]] and CSS can only be used to make static/passive documents, whilst [[JavaScript]] represents the real "threat" to be countered through disabling.{{Citation needed}}<!-- Chromium disables HTML-based lazy-loading when JS is disabled, for privacy reasons. But it doesn't disable lazy-load for CSS --> | ||
==Examples== | ==Examples== | ||