A new technique for identifying a browser instance has been revealed . The method is based on the image processing characteristics of favicons, which the website uses to determine the icons displayed in bookmarks, tabs, and other elements of the browser interface.
Browsers store Favicon images in a separate cache , which does not overlap with other caches, is common to all modes of operation, and is not cleared by standard cache and browsing history cleaners.
This feature allows the identifier to be used even when working in incognito mode and makes it difficult to remove. Authentication using this method is also unaffected by the use of VPNs and ad-blocking plugins.
The identification method is based on the fact that on the server side it is possible to determine whether the user has previously opened the page by analyzing the information about the Favicon load if the browser did not request the Favicon image specified in the parameters of the page, then the page was loaded earlier and the image is displayed from the cache.
Since browsers allow you to configure your own Favicon for each page, it is possible to encode useful information through a sequential redirection of the user to several unique pages.
The more redirects in the chain, the more identifiers can be determined (the number of identifiers is determined by the formula 2 ^ N, where N is the number of redirects). For example, 4 users can address two redirects, 3-8, 4-16, 10-1024, 24-16 million, 32-4 billion.
The disadvantage of this method is the large delays : the greater the precision, the longer it takes for redirects to open the page.
32 redirects generate identifiers for all Internet users, but cause a delay of about three seconds. For a million identifiers, the delay is approximately one and a half seconds.
The method involves working in two modes : writing and reading:
- Writing mode generates and stores an identifier for the user who first accessed the site.
- Reading mode reads a previously stored identifier.
The choice of mode depends on the request of the Favicon file for the main page of the site: if the image is requested, the data is not cached and it can be assumed that the user has not accessed the site before or the content is cached. outdated. According to the researchers, by specifying the HTTP Cache-Control header, it is possible to achieve the Favicon in the cache for up to one year.
In read-only mode, when a website is opened , the user is bound to predefined pages with their favicons. The HTTP server analyzes which favicons are requested from the server and which are displayed without accessing the server from the cache. The presence of the request is coded as "0," and its absence as "1." To ensure the identifier is preserved for future requests, a 404 error code is displayed in response to favicon requests; that is, the next time the site is opened, the browser will attempt to load these favicons again.
In write mode, in the redirect loop for pages that code "1", the correct Favicon response is returned, stored in the browser cache (when the cycle is repeated, the Favicon data will be returned from the cache, without accessing the server), and for pages that code "0" - error code 404 (if the redirect cycle is repeated, the page data will be requested again).
The method works in Chrome, Safari, Edge, and partially in Firefox. In Firefox for Linux, the use of favicons as supercookies is hampered by a feature that prevents the browser from caching the favicon.
Interestingly, the authors of the authentication method notified the Firefox developers about this feature about a year ago, noting that there was an error in the cache, but not mentioning their work and that correcting the error would lead to the possibility of user identification.
Source: https://www.cs.uic.edu