projects/torbrowser/design/index.html.en (673e5f158) - tor-webwml.git

index.html.en

<?xml version="1.0" encoding="UTF-8"?>
<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd">
<html xmlns="http://www.w3.org/1999/xhtml"><head><meta http-equiv="Content-Type" content="text/html; charset=UTF-8" /><title>The Design and Implementation of the Tor Browser [DRAFT]</title><meta name="generator" content="DocBook XSL Stylesheets V1.75.2" /></head><body><div class="article" title="The Design and Implementation of the Tor Browser [DRAFT]"><div class="titlepage"><div><div><h2 class="title"><a id="design"></a>The Design and Implementation of the Tor Browser [DRAFT]</h2></div><div><div class="author"><h3 class="author"><span class="firstname">Mike</span> <span class="surname">Perry</span></h3><div class="affiliation"><div class="address"><p><code class="email">&lt;<a class="email" href="mailto:mikeperry#torproject org">mikeperry#torproject org</a>&gt;</code></p></div></div></div></div><div><div class="author"><h3 class="author"><span class="firstname">Erinn</span> <span class="surname">Clark</span></h3><div class="affiliation"><div class="address"><p><code class="email">&lt;<a class="email" href="mailto:erinn#torproject org">erinn#torproject org</a>&gt;</code></p></div></div></div></div><div><div class="author"><h3 class="author"><span class="firstname">Steven</span> <span class="surname">Murdoch</span></h3><div class="affiliation"><div class="address"><p><code class="email">&lt;<a class="email" href="mailto:sjmurdoch#torproject org">sjmurdoch#torproject org</a>&gt;</code></p></div></div></div></div><div><p class="pubdate">Dec 28 2011</p></div></div><hr /></div><div class="toc"><p><b>Table of Contents</b></p><dl><dt><span class="sect1"><a href="#id2619754">1. Introduction</a></span></dt><dd><dl><dt><span class="sect2"><a href="#adversary">1.1. Adversary Model</a></span></dt></dl></dd><dt><span class="sect1"><a href="#DesignRequirements">2. Design Requirements and Philosophy</a></span></dt><dd><dl><dt><span class="sect2"><a href="#security">2.1. Security Requirements</a></span></dt><dt><span class="sect2"><a href="#privacy">2.2. Privacy Requirements</a></span></dt><dt><span class="sect2"><a href="#philosophy">2.3. Philosophy</a></span></dt></dl></dd><dt><span class="sect1"><a href="#Implementation">3. Implementation</a></span></dt><dd><dl><dt><span class="sect2"><a href="#proxy-obedience">3.1. Proxy Obedience</a></span></dt><dt><span class="sect2"><a href="#state-separation">3.2. State Separation</a></span></dt><dt><span class="sect2"><a href="#disk-avoidance">3.3. Disk Avoidance</a></span></dt><dt><span class="sect2"><a href="#app-data-isolation">3.4. Application Data Isolation</a></span></dt><dt><span class="sect2"><a href="#identifier-linkability">3.5. Cross-Origin Identifier Unlinkability</a></span></dt><dt><span class="sect2"><a href="#fingerprinting-linkability">3.6. Cross-Origin Fingerprinting Unlinkability</a></span></dt><dt><span class="sect2"><a href="#new-identity">3.7. Long-Term Unlinkability via "New Identity" button</a></span></dt><dt><span class="sect2"><a href="#click-to-play">3.8. Click-to-play for plugins and invasive content</a></span></dt><dt><span class="sect2"><a href="#firefox-patches">3.9. Description of Firefox Patches</a></span></dt></dl></dd><dt><span class="sect1"><a href="#Packaging">4. Packaging</a></span></dt><dd><dl><dt><span class="sect2"><a href="#build-security">4.1. Build Process Security</a></span></dt><dt><span class="sect2"><a href="#addons">4.2. External Addons</a></span></dt><dt><span class="sect2"><a href="#prefs">4.3. Pref Changes</a></span></dt><dt><span class="sect2"><a href="#update-mechanism">4.4. Update Security</a></span></dt></dl></dd><dt><span class="sect1"><a href="#Testing">5. Testing</a></span></dt><dd><dl><dt><span class="sect2"><a href="#SingleStateTesting">5.1. Single state testing</a></span></dt></dl></dd></dl></div><div class="sect1" title="1. Introduction"><div class="titlepage"><div><div><h2 class="title" style="clear: both"><a id="id2619754"></a>1. Introduction</h2></div></div></div><p>

This document describes the <a class="link" href="#adversary" title="1.1. Adversary Model">adversary model</a>,
<a class="link" href="#DesignRequirements" title="2. Design Requirements and Philosophy">design requirements</a>,
<a class="link" href="#Implementation" title="3. Implementation">implementation</a>, <a class="link" href="#Packaging" title="4. Packaging">packaging</a> and <a class="link" href="#Testing" title="5. Testing">testing
procedures</a> of the Tor Browser. It is
current as of Tor Browser 2.2.35-1 and Torbutton 1.4.5.

</p><p>

This document is also meant to serve as a set of design requirements and to
describe a reference implementation of a Private Browsing Mode that defends
against active network adversaries, in addition to the passive forensic local
adversary currently addressed by the major browsers.

</p><div class="sect2" title="1.1. Adversary Model"><div class="titlepage"><div><div><h3 class="title"><a id="adversary"></a>1.1. Adversary Model</h3></div></div></div><p>

A Tor web browser adversary has a number of goals, capabilities, and attack
types that can be used to guide us towards a set of requirements for the
Tor Browser. Let's start with the goals.

</p><div class="sect3" title="Adversary Goals"><div class="titlepage"><div><div><h4 class="title"><a id="adversarygoals"></a>Adversary Goals</h4></div></div></div><div class="orderedlist"><ol class="orderedlist" type="1"><li class="listitem"><span class="command"><strong>Bypassing proxy settings</strong></span><p>The adversary's primary goal is direct compromise and bypass of 
Tor, causing the user to directly connect to an IP of the adversary's
choosing.</p></li><li class="listitem"><span class="command"><strong>Correlation of Tor vs Non-Tor Activity</strong></span><p>If direct proxy bypass is not possible, the adversary will likely
happily settle for the ability to correlate something a user did via Tor with
their non-Tor activity. This can be done with cookies, cache identifiers,
javascript events, and even CSS. Sometimes the fact that a user uses Tor may
be enough for some authorities.</p></li><li class="listitem"><span class="command"><strong>History disclosure</strong></span><p>
The adversary may also be interested in history disclosure: the ability to
query a user's history to see if they have issued certain censored search
queries, or visited censored sites.
     </p></li><li class="listitem"><span class="command"><strong>Location information</strong></span><p>

Location information such as timezone and locality can be useful for the
adversary to determine if a user is in fact originating from one of the
regions they are attempting to control, or to zero-in on the geographical
location of a particular dissident or whistleblower.

</p></li><li class="listitem"><span class="command"><strong>Miscellaneous anonymity set reduction</strong></span><p>

Anonymity set reduction is also useful in attempting to zero in on a
particular individual. If the dissident or whistleblower is using a rare build
of Firefox for an obscure operating system, this can be very useful
information for tracking them down, or at least <a class="link" href="#fingerprinting">tracking their activities</a>.

</p></li><li class="listitem"><span class="command"><strong>History records and other on-disk
information</strong></span><p>
In some cases, the adversary may opt for a heavy-handed approach, such as
seizing the computers of all Tor users in an area (especially after narrowing
the field by the above two pieces of information). History records and cache
data are the primary goals here.
     </p></li></ol></div></div><div class="sect3" title="Adversary Capabilities - Positioning"><div class="titlepage"><div><div><h4 class="title"><a id="adversarypositioning"></a>Adversary Capabilities - Positioning</h4></div></div></div><p>
The adversary can position themselves at a number of different locations in
order to execute their attacks.
    </p><div class="orderedlist"><ol class="orderedlist" type="1"><li class="listitem"><span class="command"><strong>Exit Node or Upstream Router</strong></span><p>
The adversary can run exit nodes, or alternatively, they may control routers
upstream of exit nodes. Both of these scenarios have been observed in the
wild.
     </p></li><li class="listitem"><span class="command"><strong>Ad servers and/or Malicious Websites</strong></span><p>
The adversary can also run websites, or more likely, they can contract out
ad space from a number of different ad servers and inject content that way. For
some users, the adversary may be the ad servers themselves. It is not
inconceivable that ad servers may try to subvert or reduce a user's anonymity 
through Tor for marketing purposes.
     </p></li><li class="listitem"><span class="command"><strong>Local Network/ISP/Upstream Router</strong></span><p>
The adversary can also inject malicious content at the user's upstream router
when they have Tor disabled, in an attempt to correlate their Tor and Non-Tor
activity.
     </p></li><li class="listitem"><span class="command"><strong>Physical Access</strong></span><p>
Some users face adversaries with intermittent or constant physical access.
Users in Internet cafes, for example, face such a threat. In addition, in
countries where simply using tools like Tor is illegal, users may face
confiscation of their computer equipment for excessive Tor usage or just
general suspicion.
     </p></li></ol></div></div><div class="sect3" title="Adversary Capabilities - Attacks"><div class="titlepage"><div><div><h4 class="title"><a id="attacks"></a>Adversary Capabilities - Attacks</h4></div></div></div><p>

The adversary can perform the following attacks from a number of different 
positions to accomplish various aspects of their goals. It should be noted
that many of these attacks (especially those involving IP address leakage) are
often performed by accident by websites that simply have Javascript, dynamic 
CSS elements, and plugins. Others are performed by ad servers seeking to
correlate users' activity across different IP addresses, and still others are
performed by malicious agents on the Tor network and at national firewalls.

</p><div class="orderedlist"><ol class="orderedlist" type="1"><li class="listitem"><span class="command"><strong>Read and insert identifiers</strong></span><p>

The browser contains multiple facilities for storing identifiers that the
adversary creates for the purposes of tracking users. These identifiers are
most obviously cookies, but also include HTTP auth, DOM storage, cached
scripts and other elements with embedded identifiers, client certificates, and
even TLS Session IDs.

</p><p>

An adversary in a position to perform MITM content alteration can inject
document content elements to both read and inject cookies for arbitrary
domains. In fact, even many "SSL secured" websites are vulnerable to this sort of
<a class="ulink" href="http://seclists.org/bugtraq/2007/Aug/0070.html" target="_top">active
sidejacking</a>. In addition, the ad networks of course perform tracking
with cookies as well.

</p></li><li class="listitem"><a id="fingerprinting"></a><span class="command"><strong>Fingerprint users based on browser
attributes</strong></span><p>

There is an absurd amount of information available to websites via attributes
of the browser. This information can be used to reduce anonymity set, or even
uniquely fingerprint individual users. Fingerprinting is an intimidating
problem to attempt to tackle, especially without a metric to determine or at
least intuitively understand and estimate which features will most contribute
to linkability between visits.

</p><p>

The <a class="ulink" href="https://panopticlick.eff.org/about.php" target="_top">Panopticlick study
done</a> by the EFF uses the actual entropy - the number of identifying
bits of information encoded in browser properties - as this metric. Their
<a class="ulink" href="https://wiki.mozilla.org/Fingerprinting#Data" target="_top">result data</a>
is definitely useful, and the metric is probably the appropriate one for
determining how identifying a particular browser property is. However, some
quirks of their study means that they do not extract as much information as
they could from display information: they only use desktop resolution and do
not attempt to infer the size of toolbars. In the other direction, they may be
over-counting in some areas, as they did not compute joint entropy over
multiple attributes that may exhibit a high degree of correlation. Also, new
browser features are added regularly, so the data should not be taken as
final.

</p><p>

Despite the uncertainty, all fingerprinting attacks leverage the following
attack vectors:

</p><div class="orderedlist"><ol class="orderedlist" type="a"><li class="listitem"><span class="command"><strong>Observing Request Behavior</strong></span><p>

Properties of the user's request behavior comprise the bulk of low-hanging
fingerprinting targets. These include: User agent, Accept-* headers, pipeline
usage, and request ordering. Additionally, the use of custom filters such as
AdBlock and other privacy filters can be used to fingerprint request patterns
(as an extreme example).

</p></li><li class="listitem"><span class="command"><strong>Inserting Javascript</strong></span><p>

Javascript can reveal a lot of fingerprinting information. It provides DOM
objects such as window.screen and window.navigator to extract information
about the useragent.

Also, Javascript can be used to query the user's timezone via the
<code class="function">Date()</code> object, <a class="ulink" href="https://www.khronos.org/registry/webgl/specs/1.0/#5.13" target="_top">WebGL</a> can
reveal information about the video card in use, and high precision timing
information can be used to <a class="ulink" href="http://w2spconf.com/2011/papers/jspriv.pdf" target="_top">fingerprint the CPU and
interpreter speed</a>. In the future, new JavaScript features such as
<a class="ulink" href="http://w3c-test.org/webperf/specs/ResourceTiming/" target="_top">Resource
Timing</a> may leak an unknown amount of network timing related
information.

</p></li><li class="listitem"><span class="command"><strong>Inserting Plugins</strong></span><p>

The Panopticlick project found that the mere list of installed plugins (in
navigator.plugins) was sufficient to provide a large degree of
fingerprintability. Additionally, plugins are capable of extracting font lists,
interface addresses, and other machine information that is beyond what the
browser would normally provide to content. In addition, plugins can be used to
store unique identifiers that are more difficult to clear than standard
cookies.  <a class="ulink" href="http://epic.org/privacy/cookies/flash.html" target="_top">Flash-based
cookies</a> fall into this category, but there are likely numerous other
examples. Beyond fingerprinting, plugins are also abysmal at obeying the proxy
settings of the browser.

</p></li><li class="listitem"><span class="command"><strong>Inserting CSS</strong></span><p>

<a class="ulink" href="https://developer.mozilla.org/En/CSS/Media_queries" target="_top">CSS media
queries</a> can be inserted to gather information about the desktop size,
widget size, display type, DPI, user agent type, and other information that
was formerly available only to Javascript.

</p></li></ol></div></li><li class="listitem"><span class="command"><strong>Remotely or locally exploit browser and/or
OS</strong></span><p>

Last, but definitely not least, the adversary can exploit either general
browser vulnerabilities, plugin vulnerabilities, or OS vulnerabilities to
install malware and surveillance software. An adversary with physical access
can perform similar actions. Regrettably, this last attack capability is
outside of our ability to defend against, but it is worth mentioning for
completeness. <a class="ulink" href="http://tails.boum.org/contribute/design/" target="_top">The Tails
system</a> however can provide some limited defenses against this
adversary.

</p></li></ol></div></div></div></div><div class="sect1" title="2. Design Requirements and Philosophy"><div class="titlepage"><div><div><h2 class="title" style="clear: both"><a id="DesignRequirements"></a>2. Design Requirements and Philosophy</h2></div></div></div><p>

The Tor Browser Design Requirements are meant to describe the properties of a
Private Browsing Mode that defends against both network and local forensic
adversaries.

</p><p>

There are two main categories of requirements: <a class="link" href="#security" title="2.1. Security Requirements">Security Requirements</a>, and <a class="link" href="#privacy" title="2.2. Privacy Requirements">Privacy Requirements</a>. Security Requirements are the
minimum properties in order for a browser to be able to support Tor and
similar privacy proxies safely. Privacy requirements are the set of properties
that cause us to prefer one browser over another.

</p><p>

While we will endorse the use of browsers that meet the security requirements,
it is primarily the privacy requirements that cause us to maintain our own
browser distribution.

</p><p>

The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL
      NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED",  "MAY", and
      "OPTIONAL" in this document are to be interpreted as described in
      <a class="ulink" href="https://www.ietf.org/rfc/rfc2119.txt" target="_top">RFC 2119</a>.

</p><div class="sect2" title="2.1. Security Requirements"><div class="titlepage"><div><div><h3 class="title"><a id="security"></a>2.1. Security Requirements</h3></div></div></div><p>

The security requirements are primarily concerned with ensuring the safe use
of Tor. Violations in these properties typically result in serious risk for
the user in terms of immediate deanonymization and/or observability. With
respect to browser support, security requirements are the minimum properties
in order for Tor to support the use of a particular browser.

</p><div class="orderedlist"><ol class="orderedlist" type="1"><li class="listitem"><a class="link" href="#proxy-obedience" title="3.1. Proxy Obedience"><span class="command"><strong>Proxy
Obedience</strong></span></a><p>The browser
MUST NOT bypass Tor proxy settings for any content.</p></li><li class="listitem"><a class="link" href="#state-separation" title="3.2. State Separation"><span class="command"><strong>State
Separation</strong></span></a><p>The browser MUST NOT provide any stored state to the content window
from other browsers or other browsing modes, including shared state from
plugins, machine identifiers, and TLS session state.
</p></li><li class="listitem"><a class="link" href="#disk-avoidance" title="3.3. Disk Avoidance"><span class="command"><strong>Disk
Avoidance</strong></span></a><p>

The browser MUST NOT write any information that is derived from or that
reveals browsing activity to the disk, or store it in memory beyond the
duration of one browsing session, unless the user has explicitly opted to
store their browsing history information to disk.

</p></li><li class="listitem"><a class="link" href="#app-data-isolation" title="3.4. Application Data Isolation"><span class="command"><strong>Application Data
Isolation</strong></span></a><p>

The components involved in providing private browsing MUST be self-contained,
or MUST provide a mechanism for rapid, complete removal of all evidence of the
use of the mode. In other words, the browser MUST NOT write or cause the
operating system to write <span class="emphasis"><em>any information</em></span> about the use
of private browsing to disk outside of the application's control. The user
must be able to ensure that secure deletion of the software is sufficient to
remove evidence of the use of the software. All exceptions and shortcomings
due to operating system behavior MUST be wiped by an uninstaller. However, due
to permissions issues with access to swap, implementations MAY choose to leave
it out of scope, and/or leave it to the Operating System/platform to implement
ephemeral-keyed encrypted swap.

</p></li></ol></div></div><div class="sect2" title="2.2. Privacy Requirements"><div class="titlepage"><div><div><h3 class="title"><a id="privacy"></a>2.2. Privacy Requirements</h3></div></div></div><p>

The privacy requirements are primarily concerned with reducing linkability:
the ability for a user's activity on one site to be linked with their activity
on another site without their knowledge or explicit consent. With respect to
browser support, privacy requirements are the set of properties that cause us
to prefer one browser over another.

</p><p>

For the purposes of the unlinkability requirements of this section as well as
the descriptions in the <a class="link" href="#Implementation" title="3. Implementation">implementation
section</a>, a <span class="command"><strong>url bar origin</strong></span> means at least the
second-level DNS name.  For example, for mail.google.com, the origin would be
google.com. Implementations MAY, at their option, restrict the url bar origin
to be the entire fully qualified domain name.

</p><div class="orderedlist"><ol class="orderedlist" type="1"><li class="listitem"><a class="link" href="#identifier-linkability" title="3.5. Cross-Origin Identifier Unlinkability"><span class="command"><strong>Cross-Origin
Identifier Unlinkability</strong></span></a><p>

User activity on one url bar origin MUST NOT be linkable to their activity in
any other url bar origin by any third party automatically or without user
interaction or approval. This requirement specifically applies to linkability
from stored browser identifiers, authentication tokens, and shared state. The
requirement does not apply to linkable information the user manually submits
to sites, or due to information submitted during manual link traversal. This
functionality SHOULD NOT interfere with federated login in a substantial way.

</p></li><li class="listitem"><a class="link" href="#fingerprinting-linkability" title="3.6. Cross-Origin Fingerprinting Unlinkability"><span class="command"><strong>Cross-Origin
Fingerprinting Unlinkability</strong></span></a><p>

User activity on one url bar origin MUST NOT be linkable to their activity in
any other url bar origin by any third party. This property specifically applies to
linkability from fingerprinting browser behavior.

</p></li><li class="listitem"><a class="link" href="#new-identity" title="3.7. Long-Term Unlinkability via &quot;New Identity&quot; button"><span class="command"><strong>Long-Term
Unlinkability</strong></span></a><p>

The browser SHOULD provide an obvious, easy way to remove all of its
authentication tokens and browser state and obtain a fresh identity.
Additionally, the browser SHOULD clear linkable state by default automatically
upon browser restart, except at user option.

</p></li></ol></div></div><div class="sect2" title="2.3. Philosophy"><div class="titlepage"><div><div><h3 class="title"><a id="philosophy"></a>2.3. Philosophy</h3></div></div></div><p>

In addition to the above design requirements, the technology decisions about
Tor Browser are also guided by some philosophical positions about technology.

</p><div class="orderedlist"><ol class="orderedlist" type="1"><li class="listitem"><span class="command"><strong>Preserve existing user model</strong></span><p>

The existing way that the user expects to use a browser must be preserved. If
the user has to maintain a different mental model of how the sites they are
using behave depending on tab, browser state, or anything else that would not
normally be what they experience in their default browser, the user will
inevitably be confused. They will make mistakes and reduce their privacy as a
result. Worse, they may just stop using the browser, assuming it is broken.

</p><p>

User model breakage was one of the <a class="ulink" href="https://blog.torproject.org/blog/toggle-or-not-toggle-end-torbutton" target="_top">failures
of Torbutton</a>: Even if users managed to install everything properly,
the toggle model was too hard for the average user to understand, especially
in the face of accumulating tabs from multiple states crossed with the current
tor-state of the browser.

</p></li><li class="listitem"><span class="command"><strong>Favor the implementation mechanism least likely to
break sites</strong></span><p>

In general, we try to find solutions to privacy issues that will not induce
site breakage, though this is not always possible.

</p></li><li class="listitem"><span class="command"><strong>Plugins must be restricted</strong></span><p>

Even if plugins always properly used the browser proxy settings (which none of
them do) and could not be induced to bypass them (which all of them can), the
activities of closed-source plugins are very difficult to audit and control.
They can obtain and transmit all manner of system information to websites,
often have their own identifier storage for tracking users, and also
contribute to fingerprinting.

</p><p>

Therefore, if plugins are to be enabled in private browsing modes, they must
be restricted from running automatically on every page (via click-to-play
placeholders), and/or be sandboxed to restrict the types of system calls they
can execute. If the user decides to craft an exemption to allow a plugin to be
used, it MUST only apply to the top level url bar domain, and not to all sites,
to reduce linkability.

</p></li><li class="listitem"><span class="command"><strong>Minimize Global Privacy Options</strong></span><p>

<a class="ulink" href="https://trac.torproject.org/projects/tor/ticket/3100" target="_top">Another
failure of Torbutton</a> was (and still is) the options panel. Each option
that detectably alters browser behavior can be used as a fingerprinting tool.
Similarly, all extensions <a class="ulink" href="http://blog.chromium.org/2010/06/extensions-in-incognito.html" target="_top">SHOULD be
disabled in the mode</a> except as an opt-in basis. We SHOULD NOT load
system-wide addons or plugins.

</p><p>
Instead of global browser privacy options, privacy decisions SHOULD be made
<a class="ulink" href="https://wiki.mozilla.org/Privacy/Features/Site-based_data_management_UI" target="_top">per
url bar origin</a> to eliminate the possibility of linkability
between domains. For example, when a plugin object (or a Javascript access of
window.plugins) is present in a page, the user should be given the choice of
allowing that plugin object for that url bar origin only. The same
goes for exemptions to third party cookie policy, geo-location, and any other
privacy permissions.
     </p><p>
If the user has indicated they do not care about local history storage, these
permissions can be written to disk. Otherwise, they should remain memory-only. 
     </p></li><li class="listitem"><span class="command"><strong>No filters</strong></span><p>

Filter-based addons such as <a class="ulink" href="https://addons.mozilla.org/en-US/firefox/addon/adblock-plus/" target="_top">AdBlock
Plus</a>, <a class="ulink" href="http://requestpolicy.com/" target="_top">Request Policy</a>,
<a class="ulink" href="http://www.ghostery.com/about" target="_top">Ghostery</a>, <a class="ulink" href="http://priv3.icsi.berkeley.edu/" target="_top">Priv3</a>, and <a class="ulink" href="http://sharemenot.cs.washington.edu/" target="_top">Sharemenot</a> are to be
avoided. We believe that these addons do not add any real privacy to a proper
<a class="link" href="#Implementation" title="3. Implementation">implementation</a> of the above <a class="link" href="#privacy" title="2.2. Privacy Requirements">privacy requirements</a>, as all third parties are
prevented from tracking users between sites by the implementation.
Filter-based addons can also introduce strange breakage and cause usability
nightmares, and will also fail to do their job if an adversary simply
registers a new domain or creates a new url path. Worse still, the unique
filter sets that each user creates or installs will provide a wealth of
fingerprinting targets.

</p><p>

As a general matter, we are also generally opposed to shipping an always-on Ad
blocker with Tor Browser. We feel that this would damage our credibility in
terms of demonstrating that we are providing privacy through a sound design
alone, as well as damage the acceptance of Tor users by sites who support
themselves through advertising revenue.

</p><p>
Users are free to install these addons if they wish, but doing
so is not recommended, as it will alter the browser request fingerprint.
      </p></li><li class="listitem"><span class="command"><strong>Stay Current</strong></span><p>
We believe that if we do not stay current with the support of new web
technologies, we cannot hope to substantially influence or be involved in
their proper deployment or privacy realization. However, we will likely disable
high-risk features pending analysis, audit, and mitigation.
      </p></li></ol></div></div></div><div class="sect1" title="3. Implementation"><div class="titlepage"><div><div><h2 class="title" style="clear: both"><a id="Implementation"></a>3. Implementation</h2></div></div></div><p>

The Implementation section is divided into subsections, each of which
corresponds to a <a class="link" href="#DesignRequirements" title="2. Design Requirements and Philosophy">Design Requirement</a>.
Each subsection is divided into specific web technologies or properties. The
implementation is then described for that property.

</p><p>

In some cases, the implementation meets the design requirements in a non-ideal
way (for example, by disabling features). In rare cases, there may be no
implementation at all. Both of these cases are denoted by differentiating
between the <span class="command"><strong>Design Goal</strong></span> and the <span class="command"><strong>Implementation
Status</strong></span> for each property. Corresponding bugs in the <a class="ulink" href="https://trac.torproject.org/projects/tor/report" target="_top">Tor bug tracker</a>
are typically linked for these cases.

</p><div class="sect2" title="3.1. Proxy Obedience"><div class="titlepage"><div><div><h3 class="title"><a id="proxy-obedience"></a>3.1. Proxy Obedience</h3></div></div></div><p>

Proxy obedience is assured through the following:
   </p><div class="orderedlist"><ol class="orderedlist" type="1"><li class="listitem">Firefox Proxy settings
 <p>
  The Torbutton xpi sets the Firefox proxy settings to use Tor directly as a
SOCKS proxy. It sets <span class="command"><strong>network.proxy.socks_remote_dns</strong></span>,
<span class="command"><strong>network.proxy.socks_version</strong></span>, and
<span class="command"><strong>network.proxy.socks_port</strong></span>.
 </p><p>

We have verified that these settings properly proxy HTTPS, OCSP, HTTP, FTP,
gopher (now defunct), DNS, SafeBrowsing Queries, all javascript activity,
including HTML5 audio and video objects, addon updates, wifi geolocation
queries, searchbox queries, XPCOM addon HTTPS/HTTP activity, and live bookmark
updates. We have also verified that IPv6 connections are not attempted,
through the proxy or otherwise (Tor does not yet support IPv6). We have also
verified that external protocol helpers, such as smb urls and other custom
protocol handers are all blocked.

</p><p>

Numerous other third parties have also reviewed and <a class="link" href="#SingleStateTesting" title="5.1. Single state testing">tested</a> the proxy settings
and have provided test cases based on their work. See in particular <a class="ulink" href="http://decloak.net/" target="_top">decloak.net</a>.

</p></li><li class="listitem">Disabling plugins

<p>Plugins have the ability to make arbitrary OS system calls and  <a class="ulink" href="http://decloak.net/" target="_top">bypass proxy settings</a>. This includes
the ability to make UDP sockets and send arbitrary data independent of the
browser proxy settings.
 </p><p>
Torbutton disables plugins by using the
<span class="command"><strong>@mozilla.org/plugin/host;1</strong></span> service to mark the plugin tags
as disabled. Additionally, we set
<span class="command"><strong>plugin.disable_full_page_plugin_for_types</strong></span> to the list of
supported mime types for all currently installed plugins.
 </p><p>
In addition, to prevent any unproxied activity by plugins at load time, we
also patch the Firefox source code to <a class="ulink" href="https://gitweb.torproject.org/torbrowser.git/blob/maint-2.2:/src/current-patches/firefox/0007-Block-all-plugins-except-flash.patch" target="_top">prevent the load of any plugins except
for Flash and Gnash</a>.

</p><p>

Finally, even if the user alters their browser settings to re-enable the Flash
plugin, we have configured NoScript to provide click-to-play placeholders, so
that only desired objects will be loaded, and only after user confirmation.

</p></li><li class="listitem">External App Blocking
  <p>
External apps, if launched automatically, can be induced to load files that
perform network activity. In order to prevent this, Torbutton installs a
component to 
<a class="ulink" href="https://gitweb.torproject.org/torbutton.git/blob_plain/HEAD:/src/components/external-app-blocker.js" target="_top">
provide the user with a popup</a> whenever the browser attempts to
launch a helper app.

Additionally, due primarily to an issue with Ubuntu Unity, url-based drag and drop is
filtered by this component. Unity was pre-fetching URLs without using the
browser's proxy settings during a drag action, even if the drop was ultimately
canceled by the user.
  </p></li></ol></div></div><div class="sect2" title="3.2. State Separation"><div class="titlepage"><div><div><h3 class="title"><a id="state-separation"></a>3.2. State Separation</h3></div></div></div><p>
Tor Browser State is separated from existing browser state through use of a
custom Firefox profile. Furthermore, plugins are disabled, which prevents
Flash cookies from leaking from a pre-existing Flash directory.
   </p></div><div class="sect2" title="3.3. Disk Avoidance"><div class="titlepage"><div><div><h3 class="title"><a id="disk-avoidance"></a>3.3. Disk Avoidance</h3></div></div></div><div class="sect3" title="Design Goal:"><div class="titlepage"><div><div><h4 class="title"><a id="id2652153"></a>Design Goal:</h4></div></div></div><div class="blockquote"><blockquote class="blockquote">
Tor Browser MUST (at user option) prevent all disk records of browser activity.
The user should be able to optionally enable URL history and other history
features if they so desire. Once we <a class="ulink" href="https://trac.torproject.org/projects/tor/ticket/3100" target="_top">simplify the
preferences interface</a>, we will likely just enable Private Browsing
mode by default to handle this goal.
    </blockquote></div></div><div class="sect3" title="Implementation Status:"><div class="titlepage"><div><div><h4 class="title"><a id="id2650204"></a>Implementation Status:</h4></div></div></div><div class="blockquote"><blockquote class="blockquote">
For now, Tor Browser blocks write access to the disk through Torbutton
using several Firefox preferences.

The set of prefs is:
<span class="command"><strong>dom.storage.enabled</strong></span>,
<span class="command"><strong>browser.cache.memory.enable</strong></span>,
<span class="command"><strong>network.http.use-cache</strong></span>,
<span class="command"><strong>browser.cache.disk.enable</strong></span>,
<span class="command"><strong>browser.cache.offline.enable</strong></span>,
<span class="command"><strong>general.open_location.last_url</strong></span>,
<span class="command"><strong>places.history.enabled</strong></span>,
<span class="command"><strong>browser.formfill.enable</strong></span>,
<span class="command"><strong>signon.rememberSignons</strong></span>,
<span class="command"><strong>browser.download.manager.retention</strong></span>,
and <span class="command"><strong>network.cookie.lifetimePolicy</strong></span>.
    </blockquote></div></div><p>
In addition, three Firefox patches are needed to prevent disk writes, even if
Private Browsing Mode is enabled. We need to

<a class="ulink" href="https://gitweb.torproject.org/torbrowser.git/blob/maint-2.2:/src/current-patches/firefox/0002-Make-Permissions-Manager-memory-only.patch" target="_top">prevent
the permissions manager from recording HTTPS STS state</a>,
<a class="ulink" href="https://gitweb.torproject.org/torbrowser.git/blob/maint-2.2:/src/current-patches/firefox/0003-Make-Intermediate-Cert-Store-memory-only.patch" target="_top">prevent
intermediate SSL certificates from being recorded</a>, and
<a class="ulink" href="https://gitweb.torproject.org/torbrowser.git/blob/maint-2.2:/src/current-patches/firefox/0008-Make-content-pref-service-memory-only-clearable.patch" target="_top">prevent
the content preferences service from recording site zoom</a>.

For more details on these patches, <a class="link" href="#firefox-patches" title="3.9. Description of Firefox Patches">see the
Firefox Patches section</a>.

</p></div><div class="sect2" title="3.4. Application Data Isolation"><div class="titlepage"><div><div><h3 class="title"><a id="app-data-isolation"></a>3.4. Application Data Isolation</h3></div></div></div><p>

Tor Browser Bundle MUST NOT cause any information to be written outside of the
bundle directory. This is to ensure that the user is able to completely and
safely remove the bundle without leaving other traces of Tor usage on their
computer.

</p><p>FIXME: sjmurdoch, Erinn: explain what magic we do to satisfy this,
and/or what additional work or auditing needs to be done.
   </p></div><div class="sect2" title="3.5. Cross-Origin Identifier Unlinkability"><div class="titlepage"><div><div><h3 class="title"><a id="identifier-linkability"></a>3.5. Cross-Origin Identifier Unlinkability</h3></div></div></div><p>

The Tor Browser MUST prevent a user's activity on one site from being linked
to their activity on another site. When this goal cannot yet be met with an
existing web technology, that technology or functionality is disabled. Our
<a class="link" href="#privacy" title="2.2. Privacy Requirements">design goal</a> is to ultimately eliminate the need to disable arbitrary
technologies, and instead simply alter them in ways that allows them to
function in a backwards-compatible way while avoiding linkability. Users
should be able to use federated login of various kinds to explicitly inform
sites who they are, but that information should not transparently allow a
third party to record their activity from site to site without their prior
consent.

</p><p>

The benefit of this approach comes not only in the form of reduced
linkability, but also in terms of simplified privacy UI. If all stored browser
state and permissions become associated with the url bar origin, the six or
seven different pieces of privacy UI governing these identifiers and
permissions can become just one piece of UI. For instance, a window that lists
the url bar origin for which browser state exists, possibly with a
context-menu option to drill down into specific types of state or permissions.
An example of this simplification can be seen in Figure 1.

</p><div class="figure"><a id="id2634370"></a><p class="title"><b>Figure 1. Improving the Privacy UI</b></p><div class="figure-contents"><div class="mediaobject" align="center"><img src="CookieManagers.png" align="middle" alt="Improving the Privacy UI" /></div><div class="caption"><p></p>

On the left is the standard Firefox cookie manager. On the right is a mock-up
of how isolating identifiers to the URL bar origin might simplify the privacy
UI for all data - not just cookies. Both windows represent the set of
Cookies accumulated after visiting just five sites, but the window on the
right has the option of also representing history, DOM Storage, HTTP Auth,
search form history, login values, and so on within a context menu for each
site.

</div></div></div><br class="figure-break" /><div class="orderedlist"><ol class="orderedlist" type="1"><li class="listitem">Cookies
     <p><span class="command"><strong>Design Goal:</strong></span>

All cookies MUST be double-keyed to the url bar origin and third-party
origin. There exists a <a class="ulink" href="https://bugzilla.mozilla.org/show_bug.cgi?id=565965" target="_top">Mozilla bug</a>
that contains a prototype patch, but it lacks UI, and does not apply to modern
Firefoxes.

</p><p><span class="command"><strong>Implementation Status:</strong></span>

As a stopgap to satisfy our design requirement of unlinkability, we currently
entirely disable 3rd party cookies by setting
<span class="command"><strong>network.cookie.cookieBehavior</strong></span> to 1. We would prefer that
third party content continue to function, but we believe the requirement for 
unlinkability trumps that desire.

</p></li><li class="listitem">Cache
     <p>

Cache is isolated to the url bar origin by using a technique pioneered by
Colin Jackson et al, via their work on <a class="ulink" href="http://www.safecache.com/" target="_top">SafeCache</a>. The technique re-uses the
<a class="ulink" href="https://developer.mozilla.org/en/XPCOM_Interface_Reference/nsICachingChannel" target="_top">nsICachingChannel.cacheKey</a>
attribute that Firefox uses internally to prevent improper caching and reuse
of HTTP POST data.

</p><p>

However, to <a class="ulink" href="https://trac.torproject.org/projects/tor/ticket/3666" target="_top">increase the
security of the isolation</a> and to <a class="ulink" href="https://trac.torproject.org/projects/tor/ticket/3754" target="_top">solve conflicts
with OCSP relying the cacheKey property for reuse of POST requests</a>, we
had to <a class="ulink" href="https://gitweb.torproject.org/torbrowser.git/blob/maint-2.2:/src/current-patches/firefox/0005-Add-a-string-based-cacheKey.patch" target="_top">patch
Firefox to provide a cacheDomain cache attribute</a>. We use the fully
qualified url bar domain as input to this field.

</p><p>

Furthermore, we chose a different
isolation scheme than the Stanford implementation. First, we decoupled the
cache isolation from the third party cookie attribute. Second, we use several
mechanisms to attempt to determine the actual location attribute of the
top-level window (to obtain the url bar FQDN) used to load the page, as
opposed to relying solely on the referer property.

</p><p>

Therefore, <a class="ulink" href="http://crypto.stanford.edu/sameorigin/safecachetest.html" target="_top">the original
Stanford test cases</a> are expected to fail. Functionality can still be
verified by navigating to <a class="ulink" href="about:cache" target="_top">about:cache</a> and
viewing the key used for each cache entry. Each third party element should
have an additional "domain=string" property prepended, which will list the
FQDN that was used to source the third party element.

</p></li><li class="listitem">HTTP Auth
     <p>

HTTP authentication tokens are removed for third party elements using the
<a class="ulink" href="https://developer.mozilla.org/en/Setting_HTTP_request_headers#Observers" target="_top">http-on-modify-request
observer</a> to remove the Authorization headers to prevent <a class="ulink" href="http://jeremiahgrossman.blogspot.com/2007/04/tracking-users-without-cookies.html" target="_top">silent
linkability between domains</a>.  We also needed to <a class="ulink" href="https://gitweb.torproject.org/torbrowser.git/blob/maint-2.2:/src/current-patches/firefox/0004-Add-HTTP-auth-headers-before-the-modify-request-obse.patch" target="_top">patch
Firefox to cause the headers to get added early enough</a> to allow the
observer to modify it.

</p></li><li class="listitem">DOM Storage
     <p><span class="command"><strong>Design Goal:</strong></span>

DOM storage for third party domains MUST be isolated to the url bar origin,
to prevent linkability between sites.

</p><p><span class="command"><strong>Implementation Status:</strong></span>

Because it is isolated to third party domain as opposed to top level url bar
origin, we entirely disable DOM storage as a stopgap to ensure unlinkability.

</p></li><li class="listitem">Flash cookies
     <p><span class="command"><strong>Design Goal:</strong></span>

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 <a class="ulink" href="http://www.macromedia.com/support/documentation/en/flashplayer/help/settings_manager03.html" target="_top">Flash
settings manager</a>.

</p><p><span class="command"><strong>Implementation Status:</strong></span>

We are currently <a class="ulink" href="https://trac.torproject.org/projects/tor/ticket/3974" target="_top">having
difficulties</a> causing Flash player to use this settings
file on Windows, so Flash remains difficult to enable.

</p></li><li class="listitem">SSL+TLS session resumption and HTTP Keep-Alive
     <p><span class="command"><strong>Design Goal:</strong></span>

TLS session resumption tickets and SSL Session IDs MUST be limited to the url
bar origin.  HTTP Keep-Alive connections from a third party in one url bar
origin MUST NOT be reused for that same third party in another url bar origin.

</p><p><span class="command"><strong>Implementation Status:</strong></span>

We currently clear SSL Session IDs upon <a class="link" href="#new-identity" title="3.7. Long-Term Unlinkability via &quot;New Identity&quot; button">New
Identity</a>, we disable TLS Session Tickets via the Firefox Pref
<span class="command"><strong>security.enable_tls_session_tickets</strong></span>. We disable SSL Session
IDs via a <a class="ulink" href="https://gitweb.torproject.org/torbrowser.git/blob/maint-2.2:/src/current-patches/firefox/0010-Disable-SSL-Session-ID-tracking.patch" target="_top">patch
to Firefox</a>. To compensate for the increased round trip latency from disabling
these performance optimizations, we also enable
<a class="ulink" href="https://tools.ietf.org/html/draft-bmoeller-tls-falsestart-00" target="_top">TLS
False Start</a> via the Firefox Pref 
<span class="command"><strong>security.ssl.enable_false_start</strong></span>.
    </p><p>

Becuase of the extreme performance benefits of HTTP Keep-Alive for interactive
web apps, and because of the difficulties of conveying urlbar origin
information down into the Firefox HTTP layer, as a compromise we currently
merely reduce the HTTP Keep-Alive timeout to 20 seconds (which is measured
from the last packet read on the connection) using the Firefox preference
<span class="command"><strong>network.http.keep-alive.timeout</strong></span>.

</p></li><li class="listitem">Automated cross-origin redirects MUST NOT store identifiers
    <p><span class="command"><strong>Design Goal:</strong></span>

To prevent attacks aimed at subverting the Cross-Origin Identifier
Unlinkability <a class="link" href="#privacy" title="2.2. Privacy Requirements">privacy requirement</a>, 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.

</p><p>

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

</p><p><span class="command"><strong>Implementation status:</strong></span>

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 <a class="ulink" href="https://trac.torproject.org/projects/tor/ticket/3600" target="_top">trac bug
open</a> to implement what we can.

</p></li><li class="listitem">window.name
     <p>

<a class="ulink" href="https://developer.mozilla.org/En/DOM/Window.name" target="_top">window.name</a> 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
<a class="ulink" href="http://www.thomasfrank.se/sessionvars.html" target="_top">identifier
storage</a>.

</p><p>

In order to eliminate 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 link-driven navigation session, but as soon as the user
enters a new URL or navigates between https/http schemes, the property is cleared.

</p></li><li class="listitem">Auto form-fill
     <p>

We disable the password saving functionality in the browser as part of our
<a class="link" href="#disk-avoidance" title="3.3. Disk Avoidance">Disk Avoidance</a> requirement. However,
since users may decide to re-enable disk history records and password saving,
we also set the <a class="ulink" href="http://kb.mozillazine.org/Signon.autofillForms" target="_top">signon.autofillForms</a>
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.

</p></li><li class="listitem">HSTS supercookies
      <p>

An extreme (but not impossible) attack to mount is the creation of <a class="ulink" href="http://www.leviathansecurity.com/blog/archives/12-The-Double-Edged-Sword-of-HSTS-Persistence-and-Privacy.html" target="_top">HSTS
supercookies</a>. 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.

</p><p><span class="command"><strong>Design Goal:</strong></span>

There appears to be three options for us: 1. Disable HSTS entirely, and rely
instead on HTTPS-Everywhere to crawl and ship rules for HSTS sites. 2.
Restrict the number of HSTS-enabled third parties allowed per url bar origin.
3. Prevent third parties from storing HSTS rules. We have not yet decided upon
the best approach.

</p><p><span class="command"><strong>Implementation Status:</strong></span> Currently, HSTS state is
cleared by <a class="link" href="#new-identity" title="3.7. Long-Term Unlinkability via &quot;New Identity&quot; button">New Identity</a>, but we don't
defend against the creation of these cookies between <span class="command"><strong>New
Identity</strong></span> invocations.
      </p></li><li class="listitem">Exit node usage
     <p><span class="command"><strong>Design Goal:</strong></span>

Every distinct navigation session (as defined by a non-blank referer header)
MUST exit through a fresh Tor circuit in Tor Browser to prevent exit node
observers from linking concurrent browsing activity.

</p><p><span class="command"><strong>Implementation Status:</strong></span>

The Tor feature that supports this ability only exists in the 0.2.3.x-alpha
series. <a class="ulink" href="https://trac.torproject.org/projects/tor/ticket/3455" target="_top">Ticket
#3455</a> is the Torbutton ticket to make use of the new Tor
functionality.

</p></li></ol></div></div><div class="sect2" title="3.6. Cross-Origin Fingerprinting Unlinkability"><div class="titlepage"><div><div><h3 class="title"><a id="fingerprinting-linkability"></a>3.6. Cross-Origin Fingerprinting Unlinkability</h3></div></div></div><p>

In order to properly address the fingerprinting adversary on a technical
level, we need a metric to measure linkability of the various browser
properties beyond any stored origin-related state. <a class="ulink" href="https://panopticlick.eff.org/about.php" target="_top">The Panopticlick Project</a>
by the EFF provides us with exactly this metric. The researchers conducted a
survey of volunteers who were asked to visit an experiment page that harvested
many of the above components. They then computed the Shannon Entropy of the
resulting distribution of each of several key attributes to determine how many
bits of identifying information each attribute provided.

</p><p>

The study is not exhaustive, though. In particular, the test does not take in
all aspects of resolution information. It did not calculate the size of
widgets, window decoration, or toolbar size, which we believe may add high
amounts of entropy. It also did not measure clock offset and other time-based
fingerprints. Furthermore, as new browser features are added, this experiment
should be repeated to include them.

</p><p>

On the other hand, to avoid an infinite sinkhole, we reduce the efforts for
fingerprinting resistance by only concerning ourselves with reducing the
fingerprintable differences <span class="emphasis"><em>among</em></span> Tor Browser users. We
do not believe it is productive to concern ourselves with cross-browser
fingerprinting issues, at least not at this stage.

</p><div class="orderedlist"><ol class="orderedlist" type="1"><li class="listitem">Plugins
     <p>

Plugins add to fingerprinting risk via two main vectors: their mere presence in
window.navigator.plugins, as well as their internal functionality.

</p><p><span class="command"><strong>Design Goal:</strong></span>

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.

</p><p><span class="command"><strong>Implementation Status:</strong></span>

Currently, we entirely disable all plugins in Tor Browser. However, as a
compromise due to the popularity of Flash, we intend to <a class="ulink" href="https://trac.torproject.org/projects/tor/ticket/3974" target="_top">work
towards</a> a
click-to-play barrier using NoScript that is available only after the user has
specifically enabled plugins. Flash will be the only plugin available, and we
will ship a settings.sol file to disable Flash cookies, and to restrict P2P
features that likely bypass proxy settings.

</p></li><li class="listitem">Fonts
     <p>

According to the Panopticlick study, fonts provide the most linkability when
they are provided as an enumerable list in filesystem 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.

</p><p>

The sure-fire way to address font linkability is to ship the browser with a
font for every language, typeface, and style in use in the world, and to only
use those fonts at the exclusion of system fonts.  However, this set may be
impractically large. It is possible that a smaller <a class="ulink" href="https://secure.wikimedia.org/wikipedia/en/wiki/Unicode_typeface#List_of_Unicode_fonts" target="_top">common
subset</a> may be found that provides total coverage. However, we believe
that with strong url bar origin identifier isolation, a simpler approach can reduce the
number of bits available to the adversary while avoiding the rendering and
language issues of supporting a global font set.

</p><p><span class="command"><strong>Implementation Status:</strong></span>

We disable plugins, which prevents font enumeration. Additionally, we limit
both the number of font queries from CSS, as well as the total number of 
fonts that can be used in a document by patching Firefox. We create two prefs,
<span class="command"><strong>browser.display.max_font_attempts</strong></span> and
<span class="command"><strong>browser.display.max_font_count</strong></span> for this purpose. Once these
limits are reached, the browser behaves as if
<span class="command"><strong>browser.display.use_document_fonts</strong></span> was reached. We are
still working to determine optimal values for these prefs.

</p></li><li class="listitem">User Agent and HTTP Headers
     <p><span class="command"><strong>Design Goal:</strong></span>

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.

</p><p><span class="command"><strong>Implementation Status:</strong></span>

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
<a class="ulink" href="https://gitweb.torproject.org/torbrowser.git/blob/maint-2.2:/src/current-patches/firefox/0001-Block-Components.interfaces-lookupMethod-from-conten.patch" target="_top">remove
content script access</a> to Components.interfaces, which <a class="ulink" href="http://pseudo-flaw.net/tor/torbutton/fingerprint-firefox.html" target="_top">can be
used</a> to fingerprint OS, platform, and Firefox minor version.  </p></li><li class="listitem">Desktop resolution and CSS Media Queries
     <p>

Both CSS and Javascript have a lot of irrelevant information about the screen
resolution, usable desktop size, OS widget size, toolbar size, title bar size, and
other desktop features that are not at all relevant to rendering and serve
only to provide information for fingerprinting.

</p><p><span class="command"><strong>Design Goal:</strong></span>

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 ~5 fixed sizes based on the user's
desktop resolution.

</p><p><span class="command"><strong>Implementation Status:</strong></span>

We have implemented the above strategy for Javascript using Torbutton's <a class="ulink" href="https://gitweb.torproject.org/torbutton.git/blob/HEAD:/src/chrome/content/jshooks4.js" target="_top">JavaScript
hooks</a> as well as a window observer to <a class="ulink" href="https://gitweb.torproject.org/torbutton.git/blob/HEAD:/src/chrome/content/torbutton.js#l4002" target="_top">resize
new windows based on desktop resolution</a>. Additionally, we patch
Firefox to cause CSS Media Queries to use the client content window size
for all desktop size related media queries.

</p><p>

As far as we know, this fully satisfies our design goals for desktop
resolution information.

</p></li><li class="listitem">Timezone and clock offset
     <p><span class="command"><strong>Design Goal:</strong></span>

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 users 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.

</p><p><span class="command"><strong>Implementation Status:</strong></span>

We set the timezone using the TZ environment variable, which is supported on
all platforms. Additionally, we plan to <a class="ulink" href="https://trac.torproject.org/projects/tor/ticket/3652" target="_top">obtain a clock
offset from Tor</a>, but this won't be available until Tor 0.2.3.x is in
use.

</p></li><li class="listitem">Javascript performance fingerprinting
     <p>

<a class="ulink" href="http://w2spconf.com/2011/papers/jspriv.pdf" target="_top">Javascript performance
fingerprinting</a> is the act of profiling the performance
of various Javascript functions for the purpose of fingerprinting the
Javascript engine and the CPU.

</p><p><span class="command"><strong>Design Goal:</strong></span>

We have <a class="ulink" href="https://trac.torproject.org/projects/tor/ticket/3059" target="_top">several potential
mitigation approaches</a> 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. Our goal is to increase the
amount of time it takes to mount a successful attack. <a class="ulink" href="http://w2spconf.com/2011/papers/jspriv.pdf" target="_top">Mowery et al</a> 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.

</p><p><span class="command"><strong>Implementation Status:</strong></span>

We have no implementation as of yet.

</p></li><li class="listitem">Keystroke fingerprinting
     <p>

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

</p><p><span class="command"><strong>Design Goal:</strong></span>

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

</p><p><span class="command"><strong>Implementation Status:</strong></span>
We have no implementation as of yet.
     </p></li><li class="listitem">WebGL
     <p>

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

</p><p><span class="command"><strong>Design Goal:</strong></span>

Because of the large amount of potential fingerprinting vectors, we intend to
deploy a similar strategy against WebGL as for plugins. First, WebGL canvases
will have click-to-play placeholders, and will not run until authorized by the
user. Second, we intend to <a class="ulink" href="https://trac.torproject.org/projects/tor/ticket/3323" target="_top">obfuscate driver
information</a> by hooking
<span class="command"><strong>getParameter()</strong></span>,
<span class="command"><strong>getSupportedExtensions()</strong></span>,
<span class="command"><strong>getExtension()</strong></span>, and
<span class="command"><strong>getContextAttributes()</strong></span> to provide standard minimal,
driver-neutral information.

</p><p><span class="command"><strong>Implementation Status:</strong></span>

Currently we simply disable WebGL.

</p></li></ol></div></div><div class="sect2" title="3.7. Long-Term Unlinkability via &quot;New Identity&quot; button"><div class="titlepage"><div><div><h3 class="title"><a id="new-identity"></a>3.7. Long-Term Unlinkability via "New Identity" button</h3></div></div></div><p>
In order to avoid long-term linkability, we provide a "New Identity" context
menu option in Torbutton.
   </p><div class="sect3" title="Design Goal:"><div class="titlepage"><div><div><h4 class="title"><a id="id2637889"></a>Design Goal:</h4></div></div></div><div class="blockquote"><blockquote class="blockquote">

All linkable identifiers and browser state MUST be cleared by this feature.

</blockquote></div></div><div class="sect3" title="Implementation Status:"><div class="titlepage"><div><div><h4 class="title"><a id="id2630536"></a>Implementation Status:</h4></div></div></div><div class="blockquote"><blockquote class="blockquote">

First, Torbutton disables all open tabs and windows by tagging them and
blocking them via the nsIContentPolicy, and then closes each tab and
window. The extra step for blocking tabs is done as a precaution to ensure
that any asynchronous Javascript is in fact properly disabled. After closing
all of the windows, we then clear the following state: OCSP (by toggling
security.OCSP.enabled), cache, site-specific zoom and content preferences,
Cookies, DOM storage, safe browsing key, the Google wifi geolocation token (if
exists), HTTP auth, SSL Session IDs, HSTS state, close all remaining HTTP
keep-alive connections, and clear the last opened URL field (via the pref
general.open_location.last_url).  After clearing the browser state, we then
send the NEWNYM signal to the Tor control port to cause a new circuit to be
created.

</blockquote></div><div class="blockquote"><blockquote class="blockquote">
Additionally, the user is allowed to "protect" cookies of their choosing from
deletion during New Identity by using the Torbutton Cookie Protections UI to
protect the cookies they would like to keep across New Identity invocations.
    </blockquote></div></div></div><div class="sect2" title="3.8. Click-to-play for plugins and invasive content"><div class="titlepage"><div><div><h3 class="title"><a id="click-to-play"></a>3.8. Click-to-play for plugins and invasive content</h3></div></div></div><p>
Some content types are too invasive and/or too opaque for us to properly
eliminate their linkability properties. For these content types, we use
NoScript to provide click-to-play placeholders that do not activate the
content until the user clicks on it. This will eliminate the ability for an
adversary to use such content types to link users in a dragnet fashion across
arbitrary sites.
   </p><p>
Currently, the content types isolated in this way include Flash, WebGL, and
audio and video objects.
   </p></div><div class="sect2" title="3.9. Description of Firefox Patches"><div class="titlepage"><div><div><h3 class="title"><a id="firefox-patches"></a>3.9. Description of Firefox Patches</h3></div></div></div><p>
The set of patches we have against Firefox can be found in the <a class="ulink" href="https://gitweb.torproject.org/torbrowser.git/tree/maint-2.2:/src/current-patches/firefox" target="_top">current-patches directory of the torbrowser git repository</a>. They are:
   </p><div class="orderedlist"><ol class="orderedlist" type="1"><li class="listitem">Block Components.interfaces and Components.lookupMethod
     <p>

In order to reduce fingerprinting, we block access to these two interfaces
from content script. Components.lookupMethod can undo our <a class="ulink" href="https://gitweb.torproject.org/torbutton.git/blob/HEAD:/src/chrome/content/jshooks4.js" target="_top">Javascript
hooks</a>,
and Components.interfaces can be used for fingerprinting the platform, OS, and
Firebox version, but not much else.

</p></li><li class="listitem">Make Permissions Manager memory only
     <p>

This patch exposes a pref 'permissions.memory_only' that properly isolates the
permissions manager to memory, which is responsible for all user specified
site permissions, as well as stored <a class="ulink" href="https://secure.wikimedia.org/wikipedia/en/wiki/HTTP_Strict_Transport_Security" target="_top">HSTS</a>
policy from visited sites.

The pref does successfully clear the permissions manager memory if toggled. It
does not need to be set in prefs.js, and can be handled by Torbutton.

</p></li><li class="listitem">Make Intermediate Cert Store memory-only
     <p>

The intermediate certificate store records the intermediate SSL certificates
the browser has seen to date. Because these intermediate certificates are used 
by a limited number of domains (and in some cases, only a single domain),
the intermediate certificate store can serve as a low-resolution record of
browsing history.

</p><p><span class="command"><strong>Design Goal:</strong></span>

As an additional design goal, we would like to later alter this patch to allow this
information to be cleared from memory. The implementation does not currently
allow this.

</p></li><li class="listitem">Add HTTP auth headers before on-modify-request fires
     <p>

This patch provides a trivial modification to allow us to properly remove HTTP
auth for third parties. This patch allows us to defend against an adversary
attempting to use <a class="ulink" href="http://jeremiahgrossman.blogspot.com/2007/04/tracking-users-without-cookies.html" target="_top">HTTP
auth to silently track users between domains</a>.

</p></li><li class="listitem">Add a string-based cacheKey property for domain isolation
     <p>

To <a class="ulink" href="https://trac.torproject.org/projects/tor/ticket/3666" target="_top">increase the
security of cache isolation</a> and to <a class="ulink" href="https://trac.torproject.org/projects/tor/ticket/3754" target="_top">solve strange and
unknown conflicts with OCSP</a>, we had to <a class="ulink" href="https://gitweb.torproject.org/torbrowser.git/blob/refs/heads/maint-2.2:/src/current-patches/0005-Add-a-string-based-cacheKey.patch" target="_top">patch
Firefox to provide a cacheDomain cache attribute</a>. We use the url bar
FQDN as input to this field.

</p></li><li class="listitem">Randomize HTTP pipeline order and depth
     <p>
As an 
<a class="ulink" href="https://blog.torproject.org/blog/experimental-defense-website-traffic-fingerprinting" target="_top">experimental
defense against Website Traffic Fingerprinting</a>, we patch the standard
HTTP pipelining code to randomize the number of requests in a
pipeline, as well as their order.
     </p></li><li class="listitem">Block all plugins except flash
     <p>
We cannot use the <a class="ulink" href="http://www.oxymoronical.com/experiments/xpcomref/applications/Firefox/3.5/components/@mozilla.org/extensions/blocklist%3B1" target="_top">
@mozilla.org/extensions/blocklist;1</a> service, because we
actually want to stop plugins from ever entering the browser's process space
and/or executing code (for example, AV plugins that collect statistics/analyze
URLs, magical toolbars that phone home or "help" the user, skype buttons that
ruin our day, and censorship filters). Hence we rolled our own.
     </p></li><li class="listitem">Make content-prefs service memory only
     <p>
This patch prevents random URLs from being inserted into content-prefs.sqllite in
the profile directory as content prefs change (includes site-zoom and perhaps
other site prefs?).
     </p></li><li class="listitem">Make Tor Browser exit when not launched from Vidalia
     <p>

It turns out that on Windows 7 and later systems, the Taskbar attempts to
automatically learn the most frequent apps used by the user, and it recognizes
Tor Browser as a seperate app from Vidalia. This can cause users to try to
launch Tor Brower without Vidalia or a Tor instance running. Worse, the Tor
Browser will automatically find their default Firefox profile, and properly
connect directly without using Tor. This patch is a simple hack to cause Tor
Browser to immediately exit in this case.

</p></li><li class="listitem">Disable SSL Session ID tracking
     <p>

This patch is a simple 1-line hack to prevent SSL connections from caching
(and then later transmitting) their Session IDs. There was no preference to
govern this behavior, so we had to hack it by altering the SSL new connection
defaults.

</p></li><li class="listitem">Provide an observer event to close persistent connections
     <p>

This patch creates an observer event in the HTTP connection manager to close
all keep-alive connections that still happen to be open. This event is emitted
by the <a class="link" href="#new-identity" title="3.7. Long-Term Unlinkability via &quot;New Identity&quot; button">New Identity</a> button.

</p></li></ol></div></div></div><div class="sect1" title="4. Packaging"><div class="titlepage"><div><div><h2 class="title" style="clear: both"><a id="Packaging"></a>4. Packaging</h2></div></div></div><p> </p><div class="sect2" title="4.1. Build Process Security"><div class="titlepage"><div><div><h3 class="title"><a id="build-security"></a>4.1. Build Process Security</h3></div></div></div><p> </p></div><div class="sect2" title="4.2. External Addons"><div class="titlepage"><div><div><h3 class="title"><a id="addons"></a>4.2. External Addons</h3></div></div></div><p> </p><div class="sect3" title="Included Addons"><div class="titlepage"><div><div><h4 class="title"><a id="id2611402"></a>Included Addons</h4></div></div></div></div><div class="sect3" title="Excluded Addons"><div class="titlepage"><div><div><h4 class="title"><a id="id2611409"></a>Excluded Addons</h4></div></div></div></div><div class="sect3" title="Dangerous Addons"><div class="titlepage"><div><div><h4 class="title"><a id="id2611419"></a>Dangerous Addons</h4></div></div></div></div></div><div class="sect2" title="4.3. Pref Changes"><div class="titlepage"><div><div><h3 class="title"><a id="prefs"></a>4.3. Pref Changes</h3></div></div></div><p> </p></div><div class="sect2" title="4.4. Update Security"><div class="titlepage"><div><div><h3 class="title"><a id="update-mechanism"></a>4.4. Update Security</h3></div></div></div><p> </p></div></div><div class="sect1" title="5. Testing"><div class="titlepage"><div><div><h2 class="title" style="clear: both"><a id="Testing"></a>5. Testing</h2></div></div></div><p>

The purpose of this section is to cover all the known ways that Tor browser
security can be subverted from a penetration testing perspective. The hope
is that it will be useful both for creating a "Tor Safety Check"
page, and for developing novel tests and actively attacking Torbutton with the
goal of finding vulnerabilities in either it or the Mozilla components,
interfaces and settings upon which it relies.

</p><div class="sect2" title="5.1. Single state testing"><div class="titlepage"><div><div><h3 class="title"><a id="SingleStateTesting"></a>5.1. Single state testing</h3></div></div></div><p>

Torbutton is a complicated piece of software. During development, changes to
one component can affect a whole slough of unrelated features.  A number of
aggregated test suites exist that can be used to test for regressions in
Torbutton and to help aid in the development of Torbutton-like addons and
other privacy modifications of other browsers. Some of these test suites exist
as a single automated page, while others are a series of pages you must visit
individually. They are provided here for reference and future regression
testing, and also in the hope that some brave soul will one day decide to
combine them into a comprehensive automated test suite.

</p><div class="orderedlist"><ol class="orderedlist" type="1"><li class="listitem"><a class="ulink" href="http://decloak.net/" target="_top">Decloak.net</a><p>

Decloak.net is the canonical source of plugin and external-application based
proxy-bypass exploits. It is a fully automated test suite maintained by <a class="ulink" href="http://digitaloffense.net/" target="_top">HD Moore</a> as a service for people to
use to test their anonymity systems.

</p></li><li class="listitem"><a class="ulink" href="http://deanonymizer.com/" target="_top">Deanonymizer.com</a><p>

Deanonymizer.com is another automated test suite that tests for proxy bypass
and other information disclosure vulnerabilities. It is maintained by Kyle
Williams, the author of <a class="ulink" href="http://www.janusvm.com/" target="_top">JanusVM</a>
and <a class="ulink" href="http://www.januspa.com/" target="_top">JanusPA</a>.

</p></li><li class="listitem"><a class="ulink" href="https://ip-check.info" target="_top">JonDos
AnonTest</a><p>

The <a class="ulink" href="https://anonymous-proxy-servers.net/" target="_top">JonDos people</a> also provide an
anonymity tester. It is more focused on HTTP headers and behaviors than plugin bypass, and
points out a couple of headers Torbutton could do a better job with
obfuscating.

</p></li><li class="listitem"><a class="ulink" href="http://browserspy.dk" target="_top">Browserspy.dk</a><p>

Browserspy.dk provides a tremendous collection of browser fingerprinting and
general privacy tests. Unfortunately they are only available one page at a
time, and there is not really solid feedback on good vs bad behavior in
the test results.

</p></li><li class="listitem"><a class="ulink" href="http://analyze.privacy.net/" target="_top">Privacy
Analyzer</a><p>

The Privacy Analyzer provides a dump of all sorts of browser attributes and
settings that it detects, including some information on your original IP
address. Its page layout and lack of good vs bad test result feedback makes it
not as useful as a user-facing testing tool, but it does provide some
interesting checks in a single page.

</p></li><li class="listitem"><a class="ulink" href="http://ha.ckers.org/mr-t/" target="_top">Mr. T</a><p>

Mr. T is a collection of browser fingerprinting and deanonymization exploits
discovered by the <a class="ulink" href="http://ha.ckers.org" target="_top">ha.ckers.org</a> crew
and others. It is also not as user friendly as some of the above tests, but it
is a useful collection.

</p></li><li class="listitem">Gregory Fleischer's <a class="ulink" href="http://pseudo-flaw.net/content/tor/torbutton/" target="_top">Torbutton</a> and
<a class="ulink" href="http://pseudo-flaw.net/content/defcon/dc-17-demos/d.html" target="_top">Defcon
17</a> Test Cases
       <p>

Gregory Fleischer has been hacking and testing Firefox and Torbutton privacy
issues for the past 2 years. He has an excellent collection of all his test
cases that can be used for regression testing. In his Defcon work, he
demonstrates ways to infer Firefox version based on arcane browser properties.
We are still trying to determine the best way to address some of those test
cases.

</p></li><li class="listitem"><a class="ulink" href="https://torcheck.xenobite.eu/index.php" target="_top">Xenobite's
TorCheck Page</a><p>

This page checks to ensure you are using a valid Tor exit node and checks for
some basic browser properties related to privacy. It is not very fine-grained
or complete, but it is automated and could be turned into something useful
with a bit of work.

</p></li></ol></div><p>
    </p></div></div></div></body></html>

tor-webwml.git

Update design doc to describe CSS media query and font patches.