Design Goal: All cookies MUST be double-keyed to the URL bar origin and third-party origin.

Implementation Status: Double-keying cookies should just work by setting privacy.firstparty.isolate to true. However, we have not audited that yet and there is still the UI part missing for managing cookies in Private Browsing Mode. We therefore opted to keep third-party cookies disabled for now by setting network.cookie.cookieBehavior to 1.

Design Goal: All cache entries MUST be isolated to the URL bar domain.

Implementation Status: We isolate the content and image cache to the URL bar domain by setting privacy.firstparty.isolate to true.

Furthermore there is the CacheStorage API. That one is currently not available in Tor Browser as we do not allow third party cookies and are in Private Browsing Mode by default. As the cache entries are written to disk the CacheStorage API got disabled in that mode in Firefox, similar to how IndexedDB is handled. There are thoughts about enabling it by providing a memory-only database but that is still work in progress. But even if users are leaving the Private Browsing Mode and are enabling third party cookies the storage is isolated to the URL bar domain by privacy.firstparty.isolate set to true.

Finally, we have the asm.js cache. The cache entry of the script is (among others things, like type of CPU, build ID, source characters of the asm.js module etc.) keyed to the origin of the script. Lacking a good solution for binding it to the URL bar domain instead we decided to disable asm.js for the time being by setting javascript.options.asmjs to false. It remains to be seen whether keying the cache entry e.g. to the source characters of the asm.js module helps to avoid using it for cross-origin tracking of users. We did not investigate that yet.

HTTP Authorization headers can be used to encode silent third party tracking identifiers. To prevent this, we set privacy.firstparty.isolate to true.

DOM storage for third party domains MUST be isolated to the URL bar domain, to prevent linkability between sites. We achieve this by setting privacy.firstparty.isolate to true.

IndexedDB storage for third party domains MUST be isolated to the URL bar domain, to prevent linkability between sites. By default IndexedDB storage is disabled as Tor Browser is using Firefox's Private Browsing Mode and does not allow third party cookies. There are thoughts about enabling this API in Private Browsing Mode as well but that is still work in progress. However, if users are leaving this mode and are enabling third party cookies, isolation to the URL bar is achieved, though, by privacy.firstparty.isolate set to true.

Design Goal: Users should be able to click-to-play flash objects from trusted sites. To make this behavior unlinkable, we wish to include a settings file for all platforms that disables flash cookies using the Flash settings manager.

Implementation Status: We are currently having difficulties causing Flash player to use this settings file on Windows, so Flash remains difficult to enable.

Design Goal: TLS session resumption tickets and SSL Session IDs MUST be limited to the URL bar domain.

Implementation Status: We disable TLS Session Tickets and SSL Session IDs by setting security.ssl.disable_session_identifiers to true. To compensate for the increased round trip latency from disabling these performance optimizations, we also enable TLS False Start via the Firefox Pref security.ssl.enable_false_start. However, URL bar domain isolation should be working both for session tickets and session IDs but we have not verified that yet.

Design Goal: Tor circuits and HTTP connections from a third party in one URL bar origin MUST NOT be reused for that same third party in another URL bar origin.

Implementation Status: The isolation functionality is provided by a Torbutton component that sets the SOCKS username and password for each request. The Tor client has logic to prevent connections with different SOCKS usernames and passwords from using the same Tor circuit. Firefox has existing logic to ensure that connections with SOCKS proxies do not re-use existing HTTP Keep-Alive connections unless the proxy settings match. We extended this logic to cover SOCKS username and password authentication, providing us with HTTP Keep-Alive unlinkability.

While the vast majority of web requests adheres to the circuit and connection unlinkability requirement there are still corner cases we need to treat separately or that lack a fix altogether.

SharedWorkers are a special form of JavaScript Worker threads that have a shared scope between all threads from the same Javascript origin. They MUST be isolated to the URL bar domain. I.e. a SharedWorker launched from a third party from one URL bar domain MUST NOT have access to the objects created by that same third party loaded under another URL bar domain. This functionality is provided by setting privacy.firstparty.isolate to true.

The URL.createObjectURL API allows a site to load arbitrary content into a random UUID that is stored in the user's browser, and this content can be accessed via a URL of the form blob:UUID from any other content element anywhere on the web. While this UUID value is neither under control of the site nor predictable, it can still be used to tag a set of users that are of high interest to an adversary.

URIs created with URL.createObjectURL MUST be limited in scope to the first party URL bar domain that created them. We provide the isolation in Tor Browser by setting privacy.firstparty.isolate to true.

Design Goal: SPDY and HTTP/2 connections MUST be isolated to the URL bar domain. Furthermore, all associated means that could be used for cross-domain user tracking (alt-svc headers come to mind) MUST adhere to this design principle as well.

Implementation status: SPDY and HTTP/2 are currently disabled by setting the Firefox preferences network.http.spdy.enabled, network.http.spdy.enabled.v2, network.http.spdy.enabled.v3, network.http.spdy.enabled.v3-1, network.http.spdy.enabled.http2, network.http.spdy.enabled.http2draft, network.http.altsvc.enabled, and network.http.altsvc.oe to false.

Design Goal: To prevent attacks aimed at subverting the Cross-Origin Identifier Unlinkability privacy requirement, the browser MUST NOT store any identifiers (cookies, cache, DOM storage, HTTP auth, etc) for cross-origin redirect intermediaries that do not prompt for user input. For example, if a user clicks on a bit.ly URL that redirects to a doubleclick.net URL that finally redirects to a cnn.com URL, only cookies from cnn.com should be retained after the redirect chain completes.

Non-automated redirect chains that require user input at some step (such as federated login systems) SHOULD still allow identifiers to persist.

Implementation status: There are numerous ways for the user to be redirected, and the Firefox API support to detect each of them is poor. We have a trac bug open to implement what we can.

window.name is a magical DOM property that for some reason is allowed to retain a persistent value for the lifespan of a browser tab. It is possible to utilize this property for identifier storage.

In order to eliminate non-consensual linkability but still allow for sites that utilize this property to function, we reset the window.name property of tabs in Torbutton every time we encounter a blank Referer. This behavior allows window.name to persist for the duration of a click-driven navigation session, but as soon as the user enters a new URL or navigates between HTTPS/HTTP schemes, the property is cleared.

We disable the password saving functionality in the browser as part of our Disk Avoidance requirement. However, since users may decide to re-enable disk history records and password saving, we also set the signon.autofillForms preference to false to prevent saved values from immediately populating fields upon page load. Since JavaScript can read these values as soon as they appear, setting this preference prevents automatic linkability from stored passwords.

An extreme (but not impossible) attack to mount is the creation of HSTS supercookies. Since HSTS effectively stores one bit of information per domain name, an adversary in possession of numerous domains can use them to construct cookies based on stored HSTS state.

HPKP provides a mechanism for user tracking across domains as well. It allows abusing the requirement to provide a backup pin and the option to report a pin validation failure. In a tracking scenario every user gets a unique SHA-256 value serving as backup pin. This value is sent back after (deliberate) pin validation failures working in fact as a cookie.

Design Goal: HSTS and HPKP MUST be isolated to the URL bar domain.

Implementation Status: Currently, HSTS and HPKP state is both cleared by New Identity, but we don't defend against the creation and usage of any of these supercookies between New Identity invocations.

The BroadcastChannel API allows cross-site communication within the same origin. However, to avoid cross-origin linkability broadcast channels MUST instead be isolated to the URL bar domain.

We provide the isolation in Tor Browser by setting privacy.firstparty.isolate to true.

OCSP requests go to Certificate Authorities (CAs) to check for revoked certificates. They are sent once the browser is visiting a website via HTTPS and no cached results are available. Thus, to avoid information leaks, e.g. to exit relays, OCSP requests MUST go over the same circuit as the HTTPS request causing them and MUST therefore be isolated to the URL bar domain. The resulting cache entries MUST be bound to the URL bar domain as well. This functionality is provided by setting privacy.firstparty.isolate to true.

Design Goal: When visiting a website its favicon is fetched via a request originating from the browser itself (similar to the OCSP mechanism mentioned in the previous section). Those requests MUST be isolated to the URL bar domain.

Implementation Status: Favicon requests are isolated to the URL bar domain by setting privacy.firstparty.isolate to true. However, we need an additional Firefox patch to take care of favicons in tab list menuitems.

Much like blob URLs, mediasource: URIs and MediaStreams can be used to tag users. Therefore, mediasource: URIs and MediaStreams MUST be isolated to the URL bar domain. This functionality is provided by setting privacy.firstparty.isolate to true.

Firefox provides the feature to connect speculatively to remote hosts if that is either indicated in the HTML file (e.g. by link rel="preconnect" and rel="prefetch") or otherwise deemed beneficial.

Firefox does not support rel="prerender", and Mozilla has disabled speculative connections and rel="preconnect" usage where a proxy is used (see comment 3 in bug 18762 for further details). Explicit prefetching via the rel="prefetch" attribute is still performed, however.

All pre-loaded links and speculative connections MUST be isolated to the URL bar domain, if enabled. This includes isolating both Tor circuit use, as well as the caching and associate browser state for the prefetched resource.

For automatic speculative connects and rel="preconnect", we leave them disabled as per the Mozilla default for proxy settings. However, if enabled, speculative connects will be isolated to the proper first party Tor circuit by the same mechanism as is used for HTTP Keep-Alive. This is true for rel="prefetch" requests as well. For rel="preconnect", we set privacy.firstparty.isolate to true. This isolation makes both preconnecting and cache warming via rel="prefetch" ineffective for links to domains other than the current URL bar domain. For links to the same domain as the URL bar domain, the full cache warming benefit is obtained. As an optimization, any preconnecting to domains other than the current URL bar domain can thus be disabled (perhaps with the exception of frames), but we do not do this. We allow these requests to proceed, but we isolate them.

The Permissions API allows a website to query the status of different permissions. Although permissions are keyed to the origin, that is not enough to alleviate cross-linkability concerns: the combined permission state could work like an identifier given more and more permissions and their state being accessible under this API.

Design Goal: Permissions MUST be isolated to the URL bar domain.

Implementation Status: Right now we provide a Firefox patch that makes sure permissions are isolated to the URL bar domain.

End-user configuration details are by far the most severe threat to fingerprinting, as they will quickly provide enough information to uniquely identify a user. We believe it is essential to avoid exposing platform configuration details to website content at all costs. We also discourage excessive fine-grained customization of Tor Browser by minimizing and aggregating user-facing privacy and security options, as well as by discouraging the use of additional plugins and addons. When it is necessary to expose configuration details in the course of providing functionality, we strive to do so only on a per-site basis via site permissions, to avoid linkability.

Device and hardware characteristics can be determined in three ways: they can be reported explicitly by the browser, they can be inferred through browser functionality, or they can be extracted through statistical measurements of system performance. We are most concerned with the cases where this information is either directly reported or can be determined via a single use of an API or feature, and prefer to either alter functionality to prevent exposing the most variable aspects of these characteristics, place such features behind site permissions, or disable them entirely.

On the other hand, because statistical inference of system performance requires many iterations to achieve accuracy in the face of noise and concurrent activity, we are less concerned with this mechanism of extracting this information. We also expect that reducing the resolution of JavaScript's time sources will significantly increase the duration of execution required to extract accurate results, and thus make statistical approaches both unattractive and highly noticeable due to excessive resource consumption.

Operating system vendor and version differences permeate many different aspects of the browser. While it is possible to address these issues with some effort, the relative lack of diversity in operating systems causes us to primarily focus our efforts on passive operating system fingerprinting mechanisms at this point in time. For the purposes of protecting user anonymity, it is not strictly essential that the operating system be completely concealed, though we recognize that it is useful to reduce this differentiation ability where possible, especially for cases where the specific version of a system can be inferred.

While somewhat outside the scope of browser fingerprinting, for completeness it is important to mention that users themselves theoretically might be fingerprinted through their behavior while interacting with a website. This behavior includes e.g. keystrokes, mouse movements, click speed, and writing style. Basic vectors such as keystroke and mouse usage fingerprinting can be mitigated by altering JavaScript's notion of time. More advanced issues like writing style fingerprinting are the domain of other tools.

Due to vast differences in feature set and implementation behavior even between different (minor) versions of the same browser, browser vendor and version differences are simply not possible to conceal in any realistic way. It is only possible to minimize the differences among different installations of the same browser vendor and version. We make no effort to mimic any other major browser vendor, and in fact most of our fingerprinting defenses serve to differentiate Tor Browser users from normal Firefox users. Because of this, any study that lumps browser vendor and version differences into its analysis of the fingerprintability of a population is largely useless for evaluating either attacks or defenses. Unfortunately, this includes popular large-scale studies such as Panopticlick and Am I Unique. To gather usable data about Tor Browser's fingerprinting defenses we launched a Google Summer of Code project in 2016, called FPCentral, with the aim to provide us an own testbed. We set this up during 2017 and have it available now for further integration into our quality assurance efforts and possible research into improving our fingerprinting defenses and measuring their effectiveness.

Value spoofing can be used for simple cases where the browser provides some aspect of the user's configuration details, devices, hardware, or operating system directly to a website. It becomes less useful when the fingerprinting method relies on behavior to infer aspects of the hardware or operating system, rather than obtain them directly.

In cases where simple spoofing is not enough to properly conceal underlying device characteristics or operating system details, the underlying subsystem that provides the functionality for a feature or API may need to be modified or completely reimplemented. This is most common in cases where customizable or version-specific aspects of the user's operating system are visible through the browser's featureset or APIs, usually because the browser directly exposes OS-provided implementations of underlying features. In these cases, such OS-provided implementations must be replaced by a generic implementation, or at least modified by an implementation wrapper layer that makes effort to conceal any user-customized aspects of the system.

Virtualization is needed when simply reimplementing a feature in a different way is insufficient to fully conceal the underlying behavior. This is most common in instances of device and hardware fingerprinting, but since the notion of time can also be virtualized, virtualization also can apply to any instance where an accurate measurement of wall clock time is required for a fingerprinting vector to attain high accuracy.

In the event that reimplementation or virtualization is too expensive in terms of performance or engineering effort, and the relative expected usage of a feature is rare, site permissions can be used to prevent the usage of a feature for cross-site tracking. Unfortunately, site permissions become less effective once a feature is already widely overused and abused by many websites, since warning fatigue typically sets in for most users after just a few permission requests.

Due to the current bias in favor of invasive APIs that expose the maximum amount of platform information, some features and APIs are simply not salvageable in their current form. When such invasive features serve only a narrow domain or use case, or when there are alternate ways of accomplishing the same task, these features and/or certain aspects of their functionality may be simply removed.

The fact that randomization causes behaviors to differ slightly with every site visit makes it appealing at first glance, but this same property makes it very difficult to objectively measure its effectiveness. By contrast, an implementation that strives for uniformity is very simple to evaluate. Despite their current flaws, a properly designed version of Panopticlick or Am I Unique could report the entropy and uniqueness rates for all users of a single user agent version, without the need for complicated statistics about the variance of the measured behaviors. FPCentral is trying to achieve that for Tor Browser by providing feedback on acceptable browser properties and giving guidance on possible improvements.

Randomization (especially incomplete randomization) may also provide a false sense of security. When a fingerprinting attempt makes naive use of randomized information, a fingerprint will appear unstable, but may not actually be sufficiently randomized to impede a dedicated adversary. Sophisticated fingerprinting mechanisms may either ignore randomized information, or incorporate knowledge of the distribution and range of randomized values into the creation of a more stable fingerprint (by either removing the randomness, modeling it, or averaging it out).

While many end-user configuration details that the browser currently exposes may be safely replaced by false information, randomization of these details must be just as exhaustive as an approach that seeks to make these behaviors uniform. When confronting either strategy, the adversary can still make use of any details which have not been altered to be either sufficiently uniform or sufficiently random.

Furthermore, the randomization approach seems to break down when it is applied to deeper issues where underlying system functionality is directly exposed. In particular, it is not clear how to randomize the capabilities of hardware attached to a computer in such a way that it either convincingly behaves like other hardware, or such that the exact properties of the hardware that vary from user to user are sufficiently randomized. Similarly, truly concealing operating system version differences through randomization may require multiple reimplementations of the underlying operating system functionality to ensure that every operating system version is covered by the range of possible behaviors.

When randomization is introduced to features that affect site behavior, it can be very distracting for this behavior to change between visits of a given site. For the simplest cases, this will lead to minor visual nuisances. However, when this information affects reported functionality or hardware characteristics, sometimes a site will function one way on one visit, and another way on a subsequent visit.

Randomizing involves performance costs. This is especially true if the fingerprinting surface is large (like in a modern browser) and one needs more elaborate randomizing strategies (including randomized virtualization) to ensure that the randomization fully conceals the true behavior. Many calls to a cryptographically secure random number generator during the course of a page load will both serve to exhaust available entropy pools, as well as lead to increased computation while loading a page.

Improper randomization might introduce a new fingerprinting vector, as the process of generating the values for the fingerprintable attributes could be itself susceptible to side-channel attacks, analysis, or exploitation.

Plugins add to fingerprinting risk via two main vectors: their mere presence in window.navigator.plugins (because they are optional, end-user installed third party software), as well as their internal functionality.

Design Goal: All plugins that have not been specifically audited or sandboxed MUST be disabled. To reduce linkability potential, even sandboxed plugins SHOULD NOT be allowed to load objects until the user has clicked through a click-to-play barrier. Additionally, version information SHOULD be reduced or obfuscated until the plugin object is loaded. For Flash, we wish to provide a settings.sol file to disable Flash cookies, and to restrict P2P features that are likely to bypass proxy settings. We'd also like to restrict access to fonts and other system information (such as IP address and MAC address) in such a sandbox.

Implementation Status: Currently, we entirely disable all plugins in Tor Browser. However, as a compromise due to the popularity of Flash, we allow users to re-enable Flash, and flash objects are blocked behind a click-to-play barrier that is available only after the user has specifically enabled plugins. Flash is the only plugin available, the rest are entirely blocked from loading by the Firefox patches mentioned in the Proxy Obedience section. We also set the Firefox preference plugin.expose_full_path to false, to avoid leaking plugin installation information.

After plugins and plugin-provided information, we believe that the HTML5 Canvas is the single largest fingerprinting threat browsers face today. Studies show that the Canvas can provide an easy-access fingerprinting target: The adversary simply renders WebGL, font, and named color data to a Canvas element, extracts the image buffer, and computes a hash of that image data. Subtle differences in the video card, font packs, and even font and graphics library versions allow the adversary to produce a stable, simple, high-entropy fingerprint of a computer. In fact, the hash of the rendered image can be used almost identically to a tracking cookie by the web server.

In some sense, the canvas can be seen as the union of many other fingerprinting vectors. If WebGL is normalized through software rendering, system colors were standardized, and the browser shipped a fixed collection of fonts (see later points in this list), it might not be necessary to create a canvas permission. However, until then, to reduce the threat from this vector, we have patched Firefox to prompt before returning valid image data to the Canvas APIs, and for access to isPointInPath and related functions. Moreover, we put media streams on a canvas behind the site permission in that patch as well. If the user hasn't previously allowed the site in the URL bar to access Canvas image data, pure white image data is returned to the JavaScript APIs. Extracting canvas image data by third parties is not allowed, though.

In Firefox, by using either WebSockets or XHR, it is possible for remote content to enumerate the list of TCP ports open on 127.0.0.1, as well as on any other machines on the local network. In other browsers, this can be accomplished by DOM events on image or script tags. This open vs filtered vs closed port list can provide a very unique fingerprint of a machine, because it essentially enables the detection of many different popular third party applications and optional system services (Skype, Bitcoin, Bittorrent and other P2P software, SSH ports, SMB and related LAN services, CUPS and printer daemon config ports, mail servers, and so on). It is also possible to determine when ports are closed versus filtered/blocked (and thus probe custom firewall configuration).

In Tor Browser, we prevent access to 127.0.0.1/localhost by ensuring that even these requests are still sent by Firefox to our SOCKS proxy (ie we set network.proxy.no_proxies_on to the empty string). The local Tor client then rejects them, since it is configured to proxy for internal IP addresses by default. Access to the local network is forbidden via the same mechanism. We also disable the WebRTC API as mentioned previously, since even if it were usable over Tor, it still currently provides the local IP address and associated network information to websites. Additionally, we rip out the option to collect local IP addresses via the NetworkInfoService.

Both NTLM and SPNEGO authentication mechanisms can leak the hostname, and in some cases the current username. The only reason why these aren't a more serious problem is that they typically involve user interaction, and likely aren't an attractive vector for this reason. However, because it is not clear if certain carefully-crafted error conditions in these protocols could cause them to reveal machine information and still fail silently prior to the password prompt, these authentication mechanisms should either be disabled, or placed behind a site permission before their use. We simply disable them with a patch.

The GamePad API provides web pages with the USB device id, product id, and driver name of all connected game controllers, as well as detailed information about their capabilities.

It's our opinion that this API needs to be completely redesigned to provide an abstract notion of a game controller rather than offloading all of the complexity associated with handling specific game controller models to web content authors. For systems without a game controller, a standard controller can be virtualized through the keyboard, which will serve to both improve usability by normalizing user interaction with different games, as well as eliminate fingerprinting vectors. Barring that, this API should be behind a site permission in Private Browsing Modes. For now though, we simply disable it via the pref dom.gamepad.enabled.

According to the Panopticlick study, fonts provide the most linkability when they are available as an enumerable list in file system order, via either the Flash or Java plugins. However, it is still possible to use CSS and/or JavaScript to query for the existence of specific fonts. With a large enough pre-built list to query, a large amount of fingerprintable information may still be available, especially given that additional fonts often end up installed by third party software and for multilingual support.

Design Goal:Font-based fingerprinting MUST be rendered ineffective

Implementation Status: We investigated shipping a predefined set of fonts to all of our users allowing only those fonts to be used by websites at the exclusion of system fonts. We are currently following this approach, which has been suggested by researchers previously. This defense is available for all three supported platforms: Windows, macOS, and Linux, although the implementations vary in detail.

For Windows and macOS we use a preference, font.system.whitelist, to restrict fonts being used to those in the whitelist. This functionality is provided by setting privacy.resistFingerprinting to true. The whitelist for Windows and macOS contains both a set of Noto fonts which we bundle and fonts provided by the operating system. For Linux systems we only bundle fonts and deploy a fonts.conf file to restrict the browser to use those fonts exclusively. In addition to that we set the font.name* preferences for macOS and Linux to make sure that a given code point is always displayed with the same font. This is not guaranteed even if we bundle all the fonts Tor Browser uses as it can happen that fonts are loaded in a different order on different systems. Setting the above mentioned preferences works around this issue by specifying the font to use explicitly.

Allowing fonts provided by the operating system for Windows and macOS users is currently a compromise between fingerprintability resistance and usability concerns. We are still investigating the right balance between them and have created a ticket in our bug tracker to summarize the current state of our defense and future work that remains to be done.

Both CSS and JavaScript have access to a lot of information about the screen resolution, usable desktop size, OS widget size, toolbar size, title bar size, and OS desktop widget sizing information that are not at all relevant to rendering and serve only to provide information for fingerprinting. Since many aspects of desktop widget positioning and size are user configurable, these properties yield customized information about the computer, even beyond the monitor size.

Design Goal: Our design goal here is to reduce the resolution information down to the bare minimum required for properly rendering inside a content window. We intend to report all rendering information correctly with respect to the size and properties of the content window, but report an effective size of 0 for all border material, and also report that the desktop is only as big as the inner content window. Additionally, new browser windows are sized such that their content windows are one of a few fixed sizes based on the user's desktop resolution. In addition, to further reduce resolution-based fingerprinting, we are investigating zoom/viewport-based mechanisms that might allow us to always report the same desktop resolution regardless of the actual size of the content window, and simply scale to make up the difference. As an alternative to zoom-based solutions we are testing a different approach in our alpha series that tries to round the browser window at all times to a multiple 200x100 pixels. Regardless which solution we finally pick, until it will be available the user should also be informed that maximizing their windows can lead to fingerprintability under the current scheme.

Implementation Status: We automatically resize new browser windows to a 200x100 pixel multiple based on desktop resolution by backporting patches from bug 1330882 and setting privacy.resistfingerprinting to true. To minimize the effect of the long tail of large monitor sizes, we also cap the window size at 1000 pixels in each direction. In addition to that we set privacy.resistFingerprinting to true to use the client content window size for window.screen, and to report a window.devicePixelRatio of 1.0. Similarly, we use that preference to return content window relative points for DOM events. We also force popups to open in new tabs (via browser.link.open_newwindow.restriction), to avoid full-screen popups inferring information about the browser resolution. In addition, we prevent auto-maximizing on browser start, and inform users that maximized windows are detrimental to privacy in this mode.

Beyond simple resolution information, a large amount of so-called "Media" information is also exported to content. Even without JavaScript, CSS has access to a lot of information about the device orientation, system theme colors, and other desktop and display features that are not at all relevant to rendering and also user configurable. Most of this information comes from CSS Media Queries, but Mozilla has exposed several user and OS theme defined color values to CSS as well.

Design Goal: A website MUST NOT be able infer anything that the user has configured about their computer. Additionally, it SHOULD NOT be able to infer machine-specific details such as screen orientation or type.

Implementation Status: We set ui.use_standins_for_native_colors to true and provide a Firefox patch to report a fixed set of system colors to content window CSS, and prevent detection of font smoothing on macOS with the help of privacy.resistFingerprinting set to true. We use the same preference, too, to always report landscape-primary for the screen orientation.

WebGL is fingerprintable both through information that is exposed about the underlying driver and optimizations, as well as through performance fingerprinting.

Because of the large amount of potential fingerprinting vectors and the previously unexposed vulnerability surface, we deploy a similar strategy against WebGL as for plugins. First, WebGL Canvases have click-to-play placeholders (provided by NoScript), and do not run until authorized by the user. Second, we obfuscate driver information by setting the Firefox preferences webgl.disable-extensions, webgl.min_capability_mode, and webgl.disable-fail-if-major-performance-caveat to true which reduces the information provided by the following WebGL API calls: getParameter(), getSupportedExtensions(), and getExtension(). Furthermore, WebGL2 is disabled by setting webgl.enable-webgl2 to false. To make the minimal WebGL mode usable we additionally normalize its properties with a Firefox patch.

Another option for WebGL might be to use software-only rendering, using a library such as Mesa. The use of such a library would avoid hardware-specific rendering differences.

The MediaDevices API provides access to connected media input devices like cameras and microphones, as well as screen sharing. In particular, it allows web content to easily enumerate those devices with MediaDevices.enumerateDevices(). This relies on WebRTC being compiled in which we currently don't do. Nevertheless, we disable this feature for now as a defense-in-depth by setting media.peerconnection.enabled and media.navigator.enabled to false.

Which MIME Types are registered with an operating system depends to a great deal on the application software and/or drivers a user chose to install. Web pages can not only estimate the amount of MIME types registered by checking navigator.mimetypes.length. Rather, they are even able to test whether particular MIME types are available which can have a non-negligible impact on a user's fingerprint. We prevent both of these information leaks by setting privacy.resistfingerprinting to true.

The Web Speech API consists of two parts: SpeechSynthesis (Text-to-Speech) and SpeechRecognition (Asynchronous Speech Recognition). The latter is still disabled in Firefox. However, the former is enabled by default and there is the risk that speechSynthesis.getVoices() has access to computer-specific speech packages making them available in an enumerable fashion. Moreover, there are callbacks that would allow JavaScript to time how long a phrase takes to be "uttered". To prevent both we set media.webspeech.synth.enabled to false.

Touch events are able to reveal the absolute screen coordinates of a device which would defeat our approach to mitigate leaking the screen size as described above. In order to prevent that we implemented two defenses: first we disable the Touch API by setting dom.w3c_touch_events.enabled to false. Second, for those user that really need or want to have this API available we patched the code to give content-window related coordinates back. Furthermore, we made sure that the touch area described by Touch.radiusX, Touch.radiusY, and Touch.rotationAngle does not leak further information and Touch.force does not reveal how much pressure a user applied to the surface. That is achieved by a direct Firefox patch which reports back 1 for the first two properties and 0.0 for the two last ones.

The Battery Status API provides access to information about the system's battery charge level. From Firefox 52 on it is disabled for web content. Initially, it was possible on Linux to get a double-precision floating point value for the charge level, which means there was a large number of possible values making it almost behave like an identifier allowing to track a user cross-origin. But still after that got fixed (and on other platforms where the precision was just two significant digits anyway) the risk for tracking users remained as combined with the chargingTime and dischargingTime the possible values got estimated to be in the millions under normal conditions. We avoid all those possible issues with disabling the Battery Status API by setting dom.battery.enabled to false.

It is possible to get the system uptime of a Tor Browser user by querying the Event.timestamp property. We avoid this by setting dom.event.highrestimestamp.enabled to true. This might seem to be counterintuitive at first glance but the effect of setting that preference to true is a normalization of evt.timestamp and new Event('').timeStamp. Together with clamping the timer resolution to 100ms this provides an effective means against system uptime fingerprinting.

KeyboardEvents provide a way for a website to find out information about the keyboard layout of its visitors. In fact there are several dimensions to this fingerprinting vector. The KeyboardEvent.code property represents a physical key that can't be changed by the keyboard layout nor by the modifier state. On the other hand the KeyboardEvent.key property contains the character that is generated by that key. This is dependent on things like keyboard layout, locale and modifier keys.

Design Goal: Websites MUST NOT be able to infer any information about the keyboard of a Tor Browser user.

Implementation Status: We provide two Firefox patches that take care of spoofing KeyboardEvent.code and KeyboardEvent.keyCode by providing consensus (US-English-style) fake properties. This is achieved by hiding the user's use of the numpad, and any non-QWERTY US English keyboard. Characters from non-en-US languages are currently returning an empty KeyboardEvent.code and a KeyboardEvent.keyCode of 0. Moreover, neither Alt or Shift, or AltGr keyboard events are reported to content.

We are currently not taking the actually deployed browser locale or the locale indicated by a loaded document into account when spoofing the keyboard layout. We think that would be the right thing to do in the longer run, to mitigate possible usability issues and broken functionality on websites. Similarily to how users of non-english Tor Browser bundles right now can choose between keeping the Accept header spoofed or not they would then be able to keep a spoofed english keyboard or a spoofed one depending on the actual Tor Browser locale or language of the document.

Design Goal: All Tor Browser users MUST provide websites with an identical user agent and HTTP header set for a given request type. We omit the Firefox minor revision, and report a popular Windows platform. If the software is kept up to date, these headers should remain identical across the population even when updated.

Implementation Status: Firefox provides several options for controlling the browser user agent string which we leverage. We also set similar prefs for controlling the Accept-Language and Accept-Charset headers, which we spoof to English by default. Additionally, we remove content script access to Components.interfaces, which can be used to fingerprint OS, platform, and Firefox minor version.

Attacks based on timing side channels are nothing new in the browser context. Cache-based, cross-site timing, and pixel stealing, to name just a few, got investigated in the past. While their fingerprinting potential varies all timing-based attacks have in common that they need sufficiently fine-grained clocks.

Design Goal: Websites MUST NOT be able to fingerprint a Tor Browser user by exploiting timing-based side channels.

Implementation Status: The cleanest solution to timing-based side channels would be to get rid of them. This has been proposed in the research community. However, we remain skeptical as it does not seem to be trivial even considering just a single side channel and more and more potential side channels are showing up. Thus, we rely on disabling all possible timing sources or making them coarse-grained enough in order to render timing side channels unsuitable as a means for fingerprinting browser users.

We set dom.enable_user_timing and dom.enable_resource_timing to false to disable these explicit timing sources. Furthermore, we clamp the resolution of explicit clocks to 100ms with two Firefox patches. This includes performance.now(), new Date().getTime() , audioContext.currentTime, canvasStream.currentTime, video.currentTime, audio.currentTime, new File([], "").lastModified , new File([], "").lastModifiedDate.getTime(), animation.startTime, animation.currentTime, animation.timeline.currentTime, and document.timeline.currentTime.

While clamping the clock resolution to 100ms is a step towards mitigating timing-based side channel fingerprinting, it is by no means sufficient. It turns out that it is possible to subvert our clamping of explicit clocks by using implicit ones, e.g. extrapolating the true time by running a busy loop with a predictable operation in it. We are tracking this problem in our bug tracker and are working with the research community and Mozilla to develop and test a proper solution to this part of our defense against timing-based side channel fingerprinting risks.

Due to bugs in Firefox it is possible to detect the locale and the platform of a Tor Browser user. Moreover, it is possible to find out the extensions a user has installed. This is done by including resource:// and/or chrome:// URIs into web content, which point to resources included in Tor Browser itself or in installed extensions, and exploiting the different behavior resulting out of that: the browser raises an exception if a webpage requests a resource but the extension is not installed. This does not happen if the extension is indeed installed but the resource path does not exist.

We believe that it should be impossible for web content to extract information out of a Tor Browser user by deploying resource:// and/or chrome:// URIs. Until this is fixed in Firefox we filter resource:// and chrome:// requests done by web content denying them by default. We need a whitelist of resource:// and chrome:// URIs, though, to avoid breaking parts of Firefox. There are more than a dozen Firefox resources do not aid in fingerprinting Tor Browser users as they are not different on the platforms and in the locales we support.

In Tor Browser, we provide non-English users the option of concealing their OS and browser locale from websites. It is debatable if this should be as high of a priority as information specific to the user's computer, but for completeness, we attempt to maintain this property.

Implementation Status: We set the fallback character set to set to windows-1252 for all locales, via intl.charset.default. We also set javascript.use_us_english_locale to true to instruct the JS engine to use en-US as its internal C locale for all Date, Math, and exception handling. Additionally, we provide a patch to use an en-US label for the isindexHTML element instead of letting the label leak the browser's UI locale.

While the latency in Tor connections varies anywhere from milliseconds to a few seconds, it is still possible for the remote site to detect large differences between the user's clock and an official reference time source.

Design Goal: All Tor Browser users MUST report the same timezone to websites. Currently, we choose UTC for this purpose, although an equally valid argument could be made for EDT/EST due to the large English-speaking population density (coupled with the fact that we spoof a US English user agent). Additionally, the Tor software should detect if the user's clock is significantly divergent from the clocks of the relays that it connects to, and use this to reset the clock values used in Tor Browser to something reasonably accurate. Alternatively, the browser can obtain this clock skew via a mechanism similar to that used in tlsdate.

Implementation Status: We set the timezone to UTC with a Firefox patch using the TZ environment variable, which is supported on all platforms. Moreover, with an additional patch just needed for the Windows platform, we make sure the TZ environment variable is respected by the ICU library as well.

JavaScript performance fingerprinting is the act of profiling the performance of various JavaScript functions for the purpose of fingerprinting the JavaScript engine and the CPU.

Design Goal: We have several potential mitigation approaches to reduce the accuracy of performance fingerprinting without risking too much damage to functionality. Our current favorite is to reduce the resolution of the Event.timeStamp and the JavaScript Date() object, while also introducing jitter. We believe that JavaScript time resolution may be reduced all the way up to the second before it seriously impacts site operation. Our goal with this quantization is to increase the amount of time it takes to mount a successful attack. Mowery et al found that even with the default precision in most browsers, they required up to 120 seconds of amortization and repeated trials to get stable results from their feature set. We intend to work with the research community to establish the optimum trade-off between quantization+jitter and amortization time, as well as identify highly variable JavaScript operations. As long as these attacks take several seconds or more to execute, they are unlikely to be appealing to advertisers, and are also very likely to be noticed if deployed against a large number of people.

Implementation Status: Currently, our mitigation against performance fingerprinting is to disable Navigation Timing by setting the Firefox preference dom.enable_performance to false, and to disable the Mozilla Video Statistics API extensions by setting the preference media.video_stats.enabled to false, too.

Keystroke fingerprinting is the act of measuring key strike time and key flight time. It is seeing increasing use as a biometric.

Design Goal: We intend to rely on the same mechanisms for defeating JavaScript performance fingerprinting: timestamp quantization and jitter.

Implementation Status: We clamp keyboard event resolution to 100ms with a Firefox patch.

Modern computers have multiple physical processor cores available in their CPU. For optimum performance, native code typically attempts to run as many threads as there are cores, and navigator.hardwareConcurrency makes the number of those threads (i.e. logical processors) available to web content.

Design Goal: Websites MUST NOT be able to fingerprint a Tor Browser user taking advantage of the amount of logical processors available.

Implementation Status: We set dom.maxHardwareConcurrency to 1 to report the same amount of logical processors for everyone. However, there are probabilistic ways of determining the same information available which we are not defending against currently. Moreover, we might even want to think about a more elaborate approach defending against this fingerprinting technique by not making all users uniform but rather by following a bucket approach as we currently do in our defense against screen size exfiltration.

The Web Audio API provides several means to aid in fingerprinting users. At the simplest level it allows differentiating between users who have the API available and those who don't by checking for an AudioContext or OscillatorNode object. However, there are more bits of information that the Web Audio API reveals if audio signals generated with an OscillatorNode are processed as hardware and software differences influence those results.

We disable the Web Audio API by setting dom.webaudio.enabled to false. That has the positive side effect that it disables one of several means to perform ultrasound cross-device tracking as well, which is based on having AudioContext available.

The MediaError object allows the user agent to report errors that occurred while handling media, for instance using audio or video elements. The message property provides specific diagnostic information to help understanding the error condition. As a defense-in-depth we make sure that no information aiding in fingerprinting is leaking to websites that way by returning just an empty string.

It is possible to monitor the connection state of a browser over time with navigator.onLine. We prevent this by setting network.manage-offline-status to false.

Reader View is a Firefox feature to view web pages clutter-free and easily adjusted to own needs and preferences. To avoid fingerprintability risks we make Tor Browser users uniform by setting reader.parse-on-load.enabled to false and browser.reader.detectedFirstArticle to true. This makes sure that documents are not parsed on load as this is disabled on some devices due to memory consumption and we pretend that everybody has already been using that feature in the past.

Tor Browser is based on Firefox which is a Mozilla product. Quite naturally, Mozilla is interested in making users aware of new features and in gathering information to learn about the most pressing needs Firefox users are facing. This is often implemented by contacting Mozilla services, be it for displaying further information about a new feature or by sending (aggregated) data back for analysis. While some of those mechanisms are disabled by default on release channels (such as telemetry data) others are not. We make sure that none of those Mozilla services are contacted to avoid possible fingerprinting risks.

In particular, we disable GeoIP-based search results by setting browser.search.countryCode and browser.search.region to US and browser.search.geoip.url to the empty string. Furthermore, we disable Selfsupport and Unified Telemetry by setting browser.selfsupport.enabled and toolkit.telemetry.unified to false and we make sure no related ping is reaching Mozilla by setting datareporting.healthreport.about.reportUrlUnified to data:text/plain,. The same is done with datareporting.healthreport.about.reportUrl and the new tiles feature related browser.newtabpage.directory.ping and browser.newtabpage.directory.source preferences. browser.newtabpage.remote is set to false in this context as well, as a defense-in-depth given that this feature is already of by default. Additionally, we disable the UITour backend by setting browser.uitour.enabled to false and avoid getting Mozilla experiments installed into Tor Browser by flipping experiments.enabled to false. On the update side we prevent the browser from pinging the new Kinto service for blocklist updates as it is not used for it yet anyway. This is done by setting services.blocklist.update_enabled to false. The captive portal detection code is disabled as well as it phones home to Mozilla. We set network.captive-portal-service.enabled to false to achieve that. Unrelated to that we make sure that Mozilla does not get bothered with TLS error reports from Tor Browser users by hiding the respective checkbox with security.ssl.errorReporting.enabled set to false. And while we have the Push API disabled as there are no Service Workers available in Tor Browser yet, we remove the value for dom.push.serverURL as a defense-in-depth. Finally, we provide a patch to prevent Mozilla's websites from querying whether particular extensions are installed and what their state in Tor Browser is by using the window.navigator.AddonManager API. As a defense-in-depth the patch makes sure that not only Mozilla's websites can't get at that information but that the whitelist governing this access is empty in general.

We have Safebrowsing disabled in Tor Browser. In order to avoid pinging providers for list updates we remove the entries for browser.safebrowsing.provider.mozilla.updateURL and browser.safebrowsing.provider.mozilla.gethashURL (and the values for Google related preferences as well).

As we mentioned in the introduction of this section, OS type fingerprinting is currently considered a lower priority, due simply to the numerous ways that characteristics of the operating system type may leak into content, and the comparatively low contribution of OS to overall entropy. In particular, there are likely to be many ways to measure the differences in widget size, scrollbar size, and other rendered details on a page. Also, directly exported OS routines (such as those from the standard C math library) expose differences in their implementations through their return values.

Design Goal: We intend to reduce or eliminate OS type fingerprinting to the best extent possible, but recognize that the effort for reward on this item is not as high as other areas. The entropy on the current OS distribution is somewhere around 2 bits, which is much lower than other vectors which can also be used to fingerprint configuration and user-specific information. You can see the major areas of OS fingerprinting we're aware of using the tbb-fingerprinting-os tag on our bug tracker.

Implementation Status: At least two HTML5 features have a different implementation status across the major OS vendors and/or the underlying hardware: the Network Connection API, and the Sensor API. We disable these APIs through the Firefox preferences dom.network.enabled and device.sensors.enabled, setting both to false.