...
|
...
|
@@ -1,9 +1,9 @@
|
1
|
1
|
<?xml version="1.0" encoding="UTF-8"?>
|
2
|
|
-<!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 V1.78.1" /></head><body><div class="article"><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"><<a class="email" href="mailto:mikeperry#torproject org">mikeperry#torproject org</a>></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"><<a class="email" href="mailto:erinn#torproject org">erinn#torproject org</a>></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"><<a class="email" href="mailto:sjmurdoch#torproject org">sjmurdoch#torproject org</a>></code></p></div></div></div></div><div><p class="pubdate">May 6th, 2015</p></div></div><hr /></div><div class="toc"><p><strong>Table of Contents</strong></p><dl class="toc"><dt><span class="sect1"><a href="#idp53435264">1. Introduction</a></span></dt><dd><dl><dt><span class="sect2"><a href="#components">1.1. Browser Component Overview</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="#adversary">3. Adversary Model</a></span></dt><dd><dl><dt><span class="sect2"><a href="#adversary-goals">3.1. Adversary Goals</a></span></dt><dt><span class="sect2"><a href="#adversary-positioning">3.2. Adversary Capabilities - Positioning</a></span></dt><dt><span class="sect2"><a href="#attacks">3.3. Adversary Capabilities - Attacks</a></span></dt></dl></dd><dt><span class="sect1"><a href="#Implementation">4. Implementation</a></span></dt><dd><dl><dt><span class="sect2"><a href="#proxy-obedience">4.1. Proxy Obedience</a></span></dt><dt><span class="sect2"><a href="#state-separation">4.2. State Separation</a></span></dt><dt><span class="sect2"><a href="#disk-avoidance">4.3. Disk Avoidance</a></span></dt><dt><span class="sect2"><a href="#app-data-isolation">4.4. Application Data Isolation</a></span></dt><dt><span class="sect2"><a href="#identifier-linkability">4.5. Cross-Origin Identifier Unlinkability</a></span></dt><dt><span class="sect2"><a href="#fingerprinting-linkability">4.6. Cross-Origin Fingerprinting Unlinkability</a></span></dt><dt><span class="sect2"><a href="#new-identity">4.7. Long-Term Unlinkability via "New Identity" button</a></span></dt><dt><span class="sect2"><a href="#other-security">4.8. Other Security Measures</a></span></dt></dl></dd><dt><span class="sect1"><a href="#BuildSecurity">5. Build Security and Package Integrity</a></span></dt><dd><dl><dt><span class="sect2"><a href="#idp55327360">5.1. Achieving Binary Reproducibility</a></span></dt><dt><span class="sect2"><a href="#idp55349120">5.2. Package Signatures and Verification</a></span></dt><dt><span class="sect2"><a href="#idp55353648">5.3. Anonymous Verification</a></span></dt><dt><span class="sect2"><a href="#update-safety">5.4. Update Safety</a></span></dt></dl></dd><dt><span class="appendix"><a href="#Transparency">A. Towards Transparency in Navigation Tracking</a></span></dt><dd><dl><dt><span class="sect1"><a href="#deprecate">A.1. Deprecation Wishlist</a></span></dt><dt><span class="sect1"><a href="#idp55389664">A.2. Promising Standards</a></span></dt></dl></dd></dl></div><div class="sect1"><div class="titlepage"><div><div><h2 class="title" style="clear: both"><a id="idp53435264"></a>1. Introduction</h2></div></div></div><p>
|
|
2
|
+<!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 V1.79.1" /></head><body><div class="article"><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"><<a class="email" href="mailto:mikeperry#torproject org">mikeperry#torproject org</a>></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"><<a class="email" href="mailto:erinn#torproject org">erinn#torproject org</a>></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"><<a class="email" href="mailto:sjmurdoch#torproject org">sjmurdoch#torproject org</a>></code></p></div></div></div></div><div><div class="author"><h3 class="author"><span class="firstname">Georg</span> <span class="surname">Koppen</span></h3><div class="affiliation"><div class="address"><p><code class="email"><<a class="email" href="mailto:gk#torproject org">gk#torproject org</a>></code></p></div></div></div></div><div><p class="pubdate">March 10th, 2017</p></div></div><hr /></div><div class="toc"><p><strong>Table of Contents</strong></p><dl class="toc"><dt><span class="sect1"><a href="#idm29">1. Introduction</a></span></dt><dd><dl><dt><span class="sect2"><a href="#components">1.1. Browser Component Overview</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="#adversary">3. Adversary Model</a></span></dt><dd><dl><dt><span class="sect2"><a href="#adversary-goals">3.1. Adversary Goals</a></span></dt><dt><span class="sect2"><a href="#adversary-positioning">3.2. Adversary Capabilities - Positioning</a></span></dt><dt><span class="sect2"><a href="#attacks">3.3. Adversary Capabilities - Attacks</a></span></dt></dl></dd><dt><span class="sect1"><a href="#Implementation">4. Implementation</a></span></dt><dd><dl><dt><span class="sect2"><a href="#proxy-obedience">4.1. Proxy Obedience</a></span></dt><dt><span class="sect2"><a href="#state-separation">4.2. State Separation</a></span></dt><dt><span class="sect2"><a href="#disk-avoidance">4.3. Disk Avoidance</a></span></dt><dt><span class="sect2"><a href="#app-data-isolation">4.4. Application Data Isolation</a></span></dt><dt><span class="sect2"><a href="#identifier-linkability">4.5. Cross-Origin Identifier Unlinkability</a></span></dt><dt><span class="sect2"><a href="#fingerprinting-linkability">4.6. Cross-Origin Fingerprinting Unlinkability</a></span></dt><dt><span class="sect2"><a href="#new-identity">4.7. Long-Term Unlinkability via "New Identity" button</a></span></dt><dt><span class="sect2"><a href="#other-security">4.8. Other Security Measures</a></span></dt></dl></dd><dt><span class="sect1"><a href="#BuildSecurity">5. Build Security and Package Integrity</a></span></dt><dd><dl><dt><span class="sect2"><a href="#idm1010">5.1. Achieving Binary Reproducibility</a></span></dt><dt><span class="sect2"><a href="#idm1042">5.2. Package Signatures and Verification</a></span></dt><dt><span class="sect2"><a href="#idm1049">5.3. Anonymous Verification</a></span></dt><dt><span class="sect2"><a href="#update-safety">5.4. Update Safety</a></span></dt></dl></dd><dt><span class="appendix"><a href="#Transparency">A. Towards Transparency in Navigation Tracking</a></span></dt><dd><dl><dt><span class="sect1"><a href="#deprecate">A.1. Deprecation Wishlist</a></span></dt><dt><span class="sect1"><a href="#idm1090">A.2. Promising Standards</a></span></dt></dl></dd></dl></div><div class="sect1"><div class="titlepage"><div><div><h2 class="title" style="clear: both"><a id="idm29"></a>1. Introduction</h2></div></div></div><p>
|
3
|
3
|
|
4
|
4
|
This document describes the <a class="link" href="#adversary" title="3. Adversary Model">adversary model</a>,
|
5
|
5
|
<a class="link" href="#DesignRequirements" title="2. Design Requirements and Philosophy">design requirements</a>, and <a class="link" href="#Implementation" title="4. Implementation">implementation</a> of the Tor Browser. It is current as of Tor Browser
|
6
|
|
-4.5.
|
|
6
|
+6.5.1.
|
7
|
7
|
|
8
|
8
|
</p><p>
|
9
|
9
|
|
...
|
...
|
@@ -25,7 +25,7 @@ Support Release (ESR) Firefox branch</a>. We have a <a class="ulink" href="https
|
25
|
25
|
against this browser to enhance privacy and security. Browser behavior is
|
26
|
26
|
additionally augmented through the <a class="ulink" href="https://gitweb.torproject.org/torbutton.git/tree/" target="_top">Torbutton
|
27
|
27
|
extension</a>, though we are in the process of moving this functionality
|
28
|
|
-into direct Firefox patches. We also <a class="ulink" href="https://gitweb.torproject.org/tor-browser.git/tree/browser/app/profile/000-tor-browser.js?h=tor-browser-31.6.0esr-4.5-1" target="_top">change
|
|
28
|
+into direct Firefox patches. We also <a class="ulink" href="https://gitweb.torproject.org/tor-browser.git/tree/browser/app/profile/000-tor-browser.js?h=tor-browser-45.8.0esr-6.5-2" target="_top">change
|
29
|
29
|
a number of Firefox preferences</a> from their defaults.
|
30
|
30
|
|
31
|
31
|
</p><p>
|
...
|
...
|
@@ -38,30 +38,31 @@ Instantbird, and XULRunner.
|
38
|
38
|
|
39
|
39
|
To help protect against potential Tor Exit Node eavesdroppers, we include
|
40
|
40
|
<a class="ulink" href="https://www.eff.org/https-everywhere" target="_top">HTTPS-Everywhere</a>. To
|
41
|
|
-provide users with optional defense-in-depth against Javascript and other
|
42
|
|
-potential exploit vectors, we also include <a class="ulink" href="https://noscript.net/" target="_top">NoScript</a>. We also modify <a class="ulink" href="https://gitweb.torproject.org/builders/tor-browser-bundle.git/tree/Bundle-Data/linux/Data/Browser/profile.default/preferences/extension-overrides.js" target="_top">several
|
|
41
|
+provide users with optional defense-in-depth against JavaScript and other
|
|
42
|
+potential exploit vectors, we also include <a class="ulink" href="http://noscript.net/" target="_top">NoScript</a>. We also modify <a class="ulink" href="https://gitweb.torproject.org/builders/tor-browser-bundle.git/tree/Bundle-Data/linux/Data/Browser/profile.default/preferences/extension-overrides.js" target="_top">several
|
43
|
43
|
extension preferences</a> from their defaults.
|
44
|
44
|
|
45
|
45
|
</p><p>
|
46
|
46
|
|
47
|
47
|
To provide censorship circumvention in areas where the public Tor network is
|
48
|
48
|
blocked either by IP, or by protocol fingerprint, we include several <a class="ulink" href="https://trac.torproject.org/projects/tor/wiki/doc/AChildsGardenOfPluggableTransports" target="_top">Pluggable
|
49
|
|
-Transports</a> in the distribution. As of this writing, we include <a class="ulink" href="https://gitweb.torproject.org/pluggable-transports/obfs4.git" target="_top">Obfs4proxy</a>,
|
|
49
|
+Transports</a> in the distribution. As of this writing, we include <a class="ulink" href="https://gitweb.torproject.org/pluggable-transports/obfs4.git" target="_top">Obfs3proxy,
|
|
50
|
+Obfs4proxy, Scramblesuit</a>,
|
50
|
51
|
<a class="ulink" href="https://trac.torproject.org/projects/tor/wiki/doc/meek" target="_top">meek</a>,
|
51
|
|
-<a class="ulink" href="https://fteproxy.org/" target="_top">FTE</a>, and <a class="ulink" href="https://crypto.stanford.edu/flashproxy/" target="_top">FlashProxy</a>.
|
|
52
|
+and <a class="ulink" href="https://fteproxy.org/" target="_top">FTE</a>.
|
52
|
53
|
|
53
|
54
|
</p></div></div><div class="sect1"><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>
|
54
|
55
|
|
55
|
56
|
The Tor Browser Design Requirements are meant to describe the properties of a
|
56
|
57
|
Private Browsing Mode that defends against both network and local forensic
|
57
|
|
-adversaries.
|
|
58
|
+adversaries.
|
58
|
59
|
|
59
|
60
|
</p><p>
|
60
|
61
|
|
61
|
62
|
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
|
62
|
63
|
minimum properties in order for a browser to be able to support Tor and
|
63
|
64
|
similar privacy proxies safely. Privacy requirements are the set of properties
|
64
|
|
-that cause us to prefer one browser over another.
|
|
65
|
+that cause us to prefer one browser over another.
|
65
|
66
|
|
66
|
67
|
</p><p>
|
67
|
68
|
|
...
|
...
|
@@ -123,7 +124,7 @@ The privacy requirements are primarily concerned with reducing linkability:
|
123
|
124
|
the ability for a user's activity on one site to be linked with their activity
|
124
|
125
|
on another site without their knowledge or explicit consent. With respect to
|
125
|
126
|
browser support, privacy requirements are the set of properties that cause us
|
126
|
|
-to prefer one browser over another.
|
|
127
|
+to prefer one browser over another.
|
127
|
128
|
|
128
|
129
|
</p><p>
|
129
|
130
|
|
...
|
...
|
@@ -181,7 +182,7 @@ User model breakage was one of the <a class="ulink" href="https://blog.torprojec
|
181
|
182
|
of Torbutton</a>: Even if users managed to install everything properly,
|
182
|
183
|
the toggle model was too hard for the average user to understand, especially
|
183
|
184
|
in the face of accumulating tabs from multiple states crossed with the current
|
184
|
|
-Tor-state of the browser.
|
|
185
|
+Tor-state of the browser.
|
185
|
186
|
|
186
|
187
|
</p></li><li class="listitem"><span class="command"><strong>Favor the implementation mechanism least likely to
|
187
|
188
|
break sites</strong></span><p>
|
...
|
...
|
@@ -221,14 +222,14 @@ system-wide and/or operating system provided addons or plugins.
|
221
|
222
|
Instead of global browser privacy options, privacy decisions should be made
|
222
|
223
|
<a class="ulink" href="https://wiki.mozilla.org/Privacy/Features/Site-based_data_management_UI" target="_top">per
|
223
|
224
|
URL bar origin</a> to eliminate the possibility of linkability
|
224
|
|
-between domains. For example, when a plugin object (or a Javascript access of
|
|
225
|
+between domains. For example, when a plugin object (or a JavaScript access of
|
225
|
226
|
window.plugins) is present in a page, the user should be given the choice of
|
226
|
227
|
allowing that plugin object for that URL bar origin only. The same
|
227
|
228
|
goes for exemptions to third party cookie policy, geolocation, and any other
|
228
|
229
|
privacy permissions.
|
229
|
230
|
</p><p>
|
230
|
231
|
If the user has indicated they wish to record local history storage, these
|
231
|
|
-permissions can be written to disk. Otherwise, they should remain memory-only.
|
|
232
|
+permissions can be written to disk. Otherwise, they should remain memory-only.
|
232
|
233
|
</p></li><li class="listitem"><span class="command"><strong>No filters</strong></span><p>
|
233
|
234
|
|
234
|
235
|
Site-specific or filter-based addons such as <a class="ulink" href="https://addons.mozilla.org/en-US/firefox/addon/adblock-plus/" target="_top">AdBlock
|
...
|
...
|
@@ -239,11 +240,29 @@ avoided. We believe that these addons do not add any real privacy to a proper
|
239
|
240
|
should be focused on general solutions that prevent tracking by all
|
240
|
241
|
third parties, rather than a list of specific URLs or hosts.
|
241
|
242
|
</p><p>
|
242
|
|
-Filter-based addons can also introduce strange breakage and cause usability
|
243
|
|
-nightmares, and will also fail to do their job if an adversary simply
|
244
|
|
-registers a new domain or creates a new URL path. Worse still, the unique
|
245
|
|
-filter sets that each user creates or installs will provide a wealth of
|
246
|
|
-fingerprinting targets.
|
|
243
|
+Implementing filter-based blocking directly into the browser, such as done with
|
|
244
|
+<a class="ulink" href="http://ieee-security.org/TC/SPW2015/W2SP/papers/W2SP_2015_submission_32.pdf" target="_top">
|
|
245
|
+Firefox' Tracking Protection</a>, does not alleviate the concerns mentioned
|
|
246
|
+in the previous paragraph. There is still just a list concerned with specific
|
|
247
|
+URLs and hosts which, in this case, are
|
|
248
|
+<a class="ulink" href="https://services.disconnect.me/disconnect-plaintext.json" target="_top">
|
|
249
|
+assembled</a> by <a class="ulink" href="https://disconnect.me/trackerprotection" target="_top">
|
|
250
|
+Disconnect</a> and <a class="ulink" href="https://github.com/mozilla-services/shavar-list-exceptions" target="_top">adapted</a> by Mozilla.
|
|
251
|
+ </p><p>
|
|
252
|
+Trying to resort to <a class="ulink" href="https://jonathanmayer.org/papers_data/bau13.pdf" target="_top">filter methods based on
|
|
253
|
+machine learning</a> does not solve the problem either: they don't provide
|
|
254
|
+a general solution to the tracking problem as they are working probabilistically.
|
|
255
|
+Even with a precision rate at 99% and a false positive rate at 0.1% trackers
|
|
256
|
+would be missed and sites would be wrongly blocked.
|
|
257
|
+ </p><p>
|
|
258
|
+Filter-based solutions in general can also introduce strange breakage and cause
|
|
259
|
+usability nightmares. Coping with those easily leads to just <a class="ulink" href="https://github.com/mozilla-services/shavar-list-exceptions" target="_top">whitelisting
|
|
260
|
+</a>
|
|
261
|
+the affected domains defeating the purpose of the filter in the first place.
|
|
262
|
+Filters will also fail to do their job if an adversary simply
|
|
263
|
+registers a new domain or <a class="ulink" href="http://ieee-security.org/TC/SPW2015/W2SP/papers/W2SP_2015_submission_24.pdf" target="_top">
|
|
264
|
+creates a new URL path</a>. Worse still, the unique filter sets that each
|
|
265
|
+user creates or installs will provide a wealth of fingerprinting targets.
|
247
|
266
|
</p><p>
|
248
|
267
|
|
249
|
268
|
As a general matter, we are also generally opposed to shipping an always-on Ad
|
...
|
...
|
@@ -266,12 +285,12 @@ A Tor web browser adversary has a number of goals, capabilities, and attack
|
266
|
285
|
types that can be used to illustrate the design requirements for the
|
267
|
286
|
Tor Browser. Let's start with the goals.
|
268
|
287
|
|
269
|
|
- </p><div class="sect2"><div class="titlepage"><div><div><h3 class="title"><a id="adversary-goals"></a>3.1. Adversary Goals</h3></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
|
|
288
|
+ </p><div class="sect2"><div class="titlepage"><div><div><h3 class="title"><a id="adversary-goals"></a>3.1. Adversary Goals</h3></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
|
270
|
289
|
Tor, causing the user to directly connect to an IP of the adversary's
|
271
|
290
|
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
|
272
|
291
|
happily settle for the ability to correlate something a user did via Tor with
|
273
|
292
|
their non-Tor activity. This can be done with cookies, cache identifiers,
|
274
|
|
-Javascript events, and even CSS. Sometimes the fact that a user uses Tor may
|
|
293
|
+JavaScript events, and even CSS. Sometimes the fact that a user uses Tor may
|
275
|
294
|
be enough for some authorities.</p></li><li class="listitem"><span class="command"><strong>History disclosure</strong></span><p>
|
276
|
295
|
The adversary may also be interested in history disclosure: the ability to
|
277
|
296
|
query a user's history to see if they have issued certain censored search
|
...
|
...
|
@@ -287,7 +306,7 @@ attempt to perform this correlation without the user's explicit consent.
|
287
|
306
|
|
288
|
307
|
Fingerprinting (more generally: "anonymity set reduction") is used to attempt
|
289
|
308
|
to gather identifying information on a particular individual without the use
|
290
|
|
-of tracking identifiers. If the dissident or whistleblower's timezone is
|
|
309
|
+of tracking identifiers. If the dissident's or whistleblower's timezone is
|
291
|
310
|
available, and they are using a rare build of Firefox for an obscure operating
|
292
|
311
|
system, and they have a specific display resolution only used on one type of
|
293
|
312
|
laptop, this can be very useful information for tracking them down, or at
|
...
|
...
|
@@ -314,7 +333,7 @@ wild.
|
314
|
333
|
The adversary can also run websites, or more likely, they can contract out
|
315
|
334
|
ad space from a number of different ad servers and inject content that way. For
|
316
|
335
|
some users, the adversary may be the ad servers themselves. It is not
|
317
|
|
-inconceivable that ad servers may try to subvert or reduce a user's anonymity
|
|
336
|
+inconceivable that ad servers may try to subvert or reduce a user's anonymity
|
318
|
337
|
through Tor for marketing purposes.
|
319
|
338
|
</p></li><li class="listitem"><span class="command"><strong>Local Network/ISP/Upstream Router</strong></span><p>
|
320
|
339
|
The adversary can also inject malicious content at the user's upstream router
|
...
|
...
|
@@ -324,7 +343,7 @@ activity.
|
324
|
343
|
|
325
|
344
|
Additionally, at this position the adversary can block Tor, or attempt to
|
326
|
345
|
recognize the traffic patterns of specific web pages at the entrance to the Tor
|
327
|
|
-network.
|
|
346
|
+network.
|
328
|
347
|
|
329
|
348
|
</p></li><li class="listitem"><span class="command"><strong>Physical Access</strong></span><p>
|
330
|
349
|
Some users face adversaries with intermittent or constant physical access.
|
...
|
...
|
@@ -334,10 +353,10 @@ confiscation of their computer equipment for excessive Tor usage or just
|
334
|
353
|
general suspicion.
|
335
|
354
|
</p></li></ol></div></div><div class="sect2"><div class="titlepage"><div><div><h3 class="title"><a id="attacks"></a>3.3. Adversary Capabilities - Attacks</h3></div></div></div><p>
|
336
|
355
|
|
337
|
|
-The adversary can perform the following attacks from a number of different
|
|
356
|
+The adversary can perform the following attacks from a number of different
|
338
|
357
|
positions to accomplish various aspects of their goals. It should be noted
|
339
|
358
|
that many of these attacks (especially those involving IP address leakage) are
|
340
|
|
-often performed by accident by websites that simply have Javascript, dynamic
|
|
359
|
+often performed by accident by websites that simply have JavaScript, dynamic
|
341
|
360
|
CSS elements, and plugins. Others are performed by ad servers seeking to
|
342
|
361
|
correlate users' activity across different IP addresses, and still others are
|
343
|
362
|
performed by malicious agents on the Tor network and at national firewalls.
|
...
|
...
|
@@ -367,22 +386,21 @@ These types of attacks are attempts at subverting our <a class="link" href="#ide
|
367
|
386
|
attributes</strong></span><p>
|
368
|
387
|
|
369
|
388
|
There is an absurd amount of information available to websites via attributes
|
370
|
|
-of the browser. This information can be used to reduce anonymity set, or even
|
371
|
|
-uniquely fingerprint individual users. Attacks of this nature are typically
|
372
|
|
-aimed at tracking users across sites without their consent, in an attempt to
|
373
|
|
-subvert our <a class="link" href="#fingerprinting-linkability" title="4.6. Cross-Origin Fingerprinting Unlinkability">Cross-Origin
|
|
389
|
+of the browser. This information can be used to reduce the anonymity set, or
|
|
390
|
+even uniquely fingerprint individual users. Attacks of this nature are
|
|
391
|
+typically aimed at tracking users across sites without their consent, in an
|
|
392
|
+attempt to subvert our <a class="link" href="#fingerprinting-linkability" title="4.6. Cross-Origin Fingerprinting Unlinkability">Cross-Origin
|
374
|
393
|
Fingerprinting Unlinkability</a> and <a class="link" href="#new-identity" title="4.7. Long-Term Unlinkability via "New Identity" button">Long-Term Unlinkability</a> design requirements.
|
375
|
394
|
|
376
|
395
|
</p><p>
|
377
|
396
|
|
378
|
|
-Fingerprinting is an intimidating
|
379
|
|
-problem to attempt to tackle, especially without a metric to determine or at
|
380
|
|
-least intuitively understand and estimate which features will most contribute
|
381
|
|
-to linkability between visits.
|
|
397
|
+Fingerprinting is an intimidating problem to attempt to tackle, especially
|
|
398
|
+without a metric to determine or at least intuitively understand and estimate
|
|
399
|
+which features will most contribute to linkability between visits.
|
382
|
400
|
|
383
|
401
|
</p><p>
|
384
|
402
|
|
385
|
|
-The <a class="ulink" href="https://panopticlick.eff.org/about.php" target="_top">Panopticlick study
|
|
403
|
+The <a class="ulink" href="https://panopticlick.eff.org/about" target="_top">Panopticlick study
|
386
|
404
|
done</a> by the EFF uses the <a class="ulink" href="https://en.wikipedia.org/wiki/Entropy_%28information_theory%29" target="_top">Shannon
|
387
|
405
|
entropy</a> - the number of identifying bits of information encoded in
|
388
|
406
|
browser properties - as this metric. Their <a class="ulink" href="https://wiki.mozilla.org/Fingerprinting#Data" target="_top">result data</a> is
|
...
|
...
|
@@ -409,20 +427,25 @@ usage, and request ordering. Additionally, the use of custom filters such as
|
409
|
427
|
AdBlock and other privacy filters can be used to fingerprint request patterns
|
410
|
428
|
(as an extreme example).
|
411
|
429
|
|
412
|
|
- </p></li><li class="listitem"><span class="command"><strong>Inserting Javascript</strong></span><p>
|
|
430
|
+ </p></li><li class="listitem"><span class="command"><strong>Inserting JavaScript</strong></span><p>
|
413
|
431
|
|
414
|
|
-Javascript can reveal a lot of fingerprinting information. It provides DOM
|
|
432
|
+JavaScript can reveal a lot of fingerprinting information. It provides DOM
|
415
|
433
|
objects such as window.screen and window.navigator to extract information
|
416
|
|
-about the user agent.
|
|
434
|
+about the user agent.
|
417
|
435
|
|
418
|
|
-Also, Javascript can be used to query the user's timezone via the
|
|
436
|
+Also, JavaScript can be used to query the user's timezone via the
|
419
|
437
|
<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
|
420
|
438
|
reveal information about the video card in use, and high precision timing
|
421
|
439
|
information can be used to <a class="ulink" href="http://w2spconf.com/2011/papers/jspriv.pdf" target="_top">fingerprint the CPU and
|
422
|
|
-interpreter speed</a>. In the future, new JavaScript features such as
|
423
|
|
-<a class="ulink" href="http://w3c-test.org/webperf/specs/ResourceTiming/" target="_top">Resource
|
424
|
|
-Timing</a> may leak an unknown amount of network timing related
|
425
|
|
-information.
|
|
440
|
+interpreter speed</a>. JavaScript features such as
|
|
441
|
+<a class="ulink" href="https://www.w3.org/TR/resource-timing/" target="_top">Resource Timing</a>
|
|
442
|
+may leak an unknown amount of network timing related information. And, moreover,
|
|
443
|
+JavaScript is able to
|
|
444
|
+<a class="ulink" href="https://seclab.cs.ucsb.edu/media/uploads/papers/sp2013_cookieless.pdf" target="_top">
|
|
445
|
+extract</a>
|
|
446
|
+<a class="ulink" href="https://www.cosic.esat.kuleuven.be/fpdetective/" target="_top">available</a>
|
|
447
|
+<a class="ulink" href="https://hal.inria.fr/hal-01285470v2/document" target="_top">fonts</a> on a
|
|
448
|
+device with high precision.
|
426
|
449
|
|
427
|
450
|
</p></li><li class="listitem"><span class="command"><strong>Inserting Plugins</strong></span><p>
|
428
|
451
|
|
...
|
...
|
@@ -435,7 +458,7 @@ store unique identifiers that are more difficult to clear than standard
|
435
|
458
|
cookies. <a class="ulink" href="http://epic.org/privacy/cookies/flash.html" target="_top">Flash-based
|
436
|
459
|
cookies</a> fall into this category, but there are likely numerous other
|
437
|
460
|
examples. Beyond fingerprinting, plugins are also abysmal at obeying the proxy
|
438
|
|
-settings of the browser.
|
|
461
|
+settings of the browser.
|
439
|
462
|
|
440
|
463
|
|
441
|
464
|
</p></li><li class="listitem"><span class="command"><strong>Inserting CSS</strong></span><p>
|
...
|
...
|
@@ -443,7 +466,7 @@ settings of the browser.
|
443
|
466
|
<a class="ulink" href="https://developer.mozilla.org/En/CSS/Media_queries" target="_top">CSS media
|
444
|
467
|
queries</a> can be inserted to gather information about the desktop size,
|
445
|
468
|
widget size, display type, DPI, user agent type, and other information that
|
446
|
|
-was formerly available only to Javascript.
|
|
469
|
+was formerly available only to JavaScript.
|
447
|
470
|
|
448
|
471
|
</p></li></ol></div></li><li class="listitem"><a id="website-traffic-fingerprinting"></a><span class="command"><strong>Website traffic fingerprinting</strong></span><p>
|
449
|
472
|
|
...
|
...
|
@@ -454,15 +477,16 @@ node itself.
|
454
|
477
|
</p><p> The most comprehensive study of the statistical properties of this
|
455
|
478
|
attack against Tor was done by <a class="ulink" href="http://lorre.uni.lu/~andriy/papers/acmccs-wpes11-fingerprinting.pdf" target="_top">Panchenko
|
456
|
479
|
et al</a>. Unfortunately, the publication bias in academia has encouraged
|
457
|
|
-the production of a number of follow-on attack papers claiming "improved"
|
458
|
|
-success rates, in some cases even claiming to completely invalidate any
|
459
|
|
-attempt at defense. These "improvements" are actually enabled primarily by
|
460
|
|
-taking a number of shortcuts (such as classifying only very small numbers of
|
461
|
|
-web pages, neglecting to publish ROC curves or at least false positive rates,
|
462
|
|
-and/or omitting the effects of dataset size on their results). Despite these
|
463
|
|
-subsequent "improvements", we are skeptical of the efficacy of this attack in
|
464
|
|
-a real world scenario, <span class="emphasis"><em>especially</em></span> in the face of any
|
465
|
|
-defenses.
|
|
480
|
+the production of
|
|
481
|
+<a class="ulink" href="https://blog.torproject.org/blog/critique-website-traffic-fingerprinting-attacks" target="_top">a
|
|
482
|
+number of follow-on attack papers claiming "improved" success rates</a>, in
|
|
483
|
+some cases even claiming to completely invalidate any attempt at defense. These
|
|
484
|
+"improvements" are actually enabled primarily by taking a number of shortcuts
|
|
485
|
+(such as classifying only very small numbers of web pages, neglecting to publish
|
|
486
|
+ROC curves or at least false positive rates, and/or omitting the effects of
|
|
487
|
+dataset size on their results). Despite these subsequent "improvements", we are
|
|
488
|
+skeptical of the efficacy of this attack in a real world scenario,
|
|
489
|
+<span class="emphasis"><em>especially</em></span> in the face of any defenses.
|
466
|
490
|
|
467
|
491
|
</p><p>
|
468
|
492
|
|
...
|
...
|
@@ -482,8 +506,8 @@ function of the complexity of the categories</a> you need to classify.
|
482
|
506
|
In the case of this attack, the key factors that increase the classification
|
483
|
507
|
complexity (and thus hinder a real world adversary who attempts this attack)
|
484
|
508
|
are large numbers of dynamically generated pages, partially cached content,
|
485
|
|
-and also the non-web activity of entire Tor network. This yields an effective
|
486
|
|
-number of "web pages" many orders of magnitude larger than even <a class="ulink" href="http://lorre.uni.lu/~andriy/papers/acmccs-wpes11-fingerprinting.pdf" target="_top">Panchenko's
|
|
509
|
+and also the non-web activity of the entire Tor network. This yields an
|
|
510
|
+effective number of "web pages" many orders of magnitude larger than even <a class="ulink" href="http://lorre.uni.lu/~andriy/papers/acmccs-wpes11-fingerprinting.pdf" target="_top">Panchenko's
|
487
|
511
|
"Open World" scenario</a>, which suffered continuous near-constant decline
|
488
|
512
|
in the true positive rate as the "Open World" size grew (see figure 4). This
|
489
|
513
|
large level of classification complexity is further confounded by a noisy and
|
...
|
...
|
@@ -522,14 +546,14 @@ can perform similar actions.
|
522
|
546
|
|
523
|
547
|
For the purposes of the browser itself, we limit the scope of this adversary
|
524
|
548
|
to one that has passive forensic access to the disk after browsing activity
|
525
|
|
-has taken place. This adversary motivates our
|
|
549
|
+has taken place. This adversary motivates our
|
526
|
550
|
<a class="link" href="#disk-avoidance" title="4.3. Disk Avoidance">Disk Avoidance</a> defenses.
|
527
|
551
|
|
528
|
552
|
</p><p>
|
529
|
553
|
|
530
|
554
|
An adversary with arbitrary code execution typically has more power, though.
|
531
|
555
|
It can be quite hard to really significantly limit the capabilities of such an
|
532
|
|
-adversary. <a class="ulink" href="http://tails.boum.org/contribute/design/" target="_top">The Tails system</a> can
|
|
556
|
+adversary. <a class="ulink" href="https://tails.boum.org/contribute/design/" target="_top">The Tails system</a> can
|
533
|
557
|
provide some defense against this adversary through the use of readonly media
|
534
|
558
|
and frequent reboots, but even this can be circumvented on machines without
|
535
|
559
|
Secure Boot through the use of BIOS rootkits.
|
...
|
...
|
@@ -555,7 +579,7 @@ are typically linked for these cases.
|
555
|
579
|
Proxy obedience is assured through the following:
|
556
|
580
|
</p><div class="orderedlist"><ol class="orderedlist" type="1"><li class="listitem"><span class="command"><strong>Firefox proxy settings, patches, and build flags</strong></span><p>
|
557
|
581
|
|
558
|
|
-Our <a class="ulink" href="https://gitweb.torproject.org/tor-browser.git/tree/browser/app/profile/000-tor-browser.js?h=tor-browser-31.6.0esr-4.5-1" target="_top">Firefox
|
|
582
|
+Our <a class="ulink" href="https://gitweb.torproject.org/tor-browser.git/tree/browser/app/profile/000-tor-browser.js?h=tor-browser-45.8.0esr-6.5-2" target="_top">Firefox
|
559
|
583
|
preferences file</a> sets the Firefox proxy settings to use Tor directly
|
560
|
584
|
as a SOCKS proxy. It sets <span class="command"><strong>network.proxy.socks_remote_dns</strong></span>,
|
561
|
585
|
<span class="command"><strong>network.proxy.socks_version</strong></span>,
|
...
|
...
|
@@ -571,9 +595,11 @@ as set the pref <span class="command"><strong>media.peerconnection.enabled</stro
|
571
|
595
|
</p><p>
|
572
|
596
|
|
573
|
597
|
We also patch Firefox in order to provide several defense-in-depth mechanisms
|
574
|
|
-for proxy safety. Notably, we <a class="ulink" href="https://gitweb.torproject.org/tor-browser.git/commit/?h=tor-browser-31.6.0esr-4.5-1&id=8c6604d2b776f0d8e33ed9130c5f5b8cf744bac8" target="_top">patch
|
|
598
|
+for proxy safety. Notably, we <a class="ulink" href="https://gitweb.torproject.org/tor-browser.git/commit/?h=tor-browser-45.8.0esr-6.5-2&id=177e78923b3252a7442160486ec48252a6adb77a" target="_top">patch
|
575
|
599
|
the DNS service</a> to prevent any browser or addon DNS resolution, and we
|
576
|
|
-also <a class="ulink" href="https://gitweb.torproject.org/tor-browser.git/commit/?h=tor-browser-31.6.0esr-4.5-1&id=c96c854c0eca21fed1362d1ddd164b657d351795" target="_top">patch
|
|
600
|
+also <a class="ulink" href="https://gitweb.torproject.org/tor-browser.git/commit/?h=tor-browser-45.8.0esr-6.5-2&id=6e17cef8f3cf61fdabf99e40d5e09a730142d6cd" target="_top">
|
|
601
|
+remove the DNS lookup for the profile lock signature</a>. Furhermore, we
|
|
602
|
+<a class="ulink" href="https://gitweb.torproject.org/tor-browser.git/commit/?h=tor-browser-45.8.0esr-6.5-2&id=8197f6ffe58ba167e3bca4230c5721ebcfae55de" target="_top">patch
|
577
|
603
|
OCSP and PKIX code</a> to prevent any use of the non-proxied command-line
|
578
|
604
|
tool utility functions from being functional while linked in to the browser.
|
579
|
605
|
In both cases, we could find no direct paths to these routines in the browser,
|
...
|
...
|
@@ -581,6 +607,36 @@ but it seemed better safe than sorry.
|
581
|
607
|
|
582
|
608
|
</p><p>
|
583
|
609
|
|
|
610
|
+For further defense-in-depth we disabled WebIDE because it can bypass proxy
|
|
611
|
+settings for remote debugging, and also because it downloads extensions we
|
|
612
|
+have not reviewed. We
|
|
613
|
+are doing this by setting
|
|
614
|
+<span class="command"><strong>devtools.webide.autoinstallADBHelper</strong></span>,
|
|
615
|
+<span class="command"><strong>devtools.webide.autoinstallFxdtAdapters</strong></span>,
|
|
616
|
+<span class="command"><strong>devtools.webide.enabled</strong></span>, and
|
|
617
|
+<span class="command"><strong>devtools.appmanager.enabled</strong></span> to <span class="command"><strong>false</strong></span>.
|
|
618
|
+Moreover, we removed the Roku Screen Sharing and screencaster code with a
|
|
619
|
+<a class="ulink" href="https://gitweb.torproject.org/tor-browser.git/commit/?h=tor-browser-45.8.0esr-6.5-2&id= ad4abdb2e724fec060063f460604b829c66ea08a" target="_top">
|
|
620
|
+Firefox patch</a> as these features can bypass proxy settings as well.
|
|
621
|
+ </p><p>
|
|
622
|
+Shumway is removed, too, for possible proxy bypass risks. We did this by
|
|
623
|
+backporting a <a class="ulink" href="https://gitweb.torproject.org/tor-browser.git/commit/?h=tor-browser-45.8.0esr-6.5-2&id=d020a4992d8d25baf7dfb5c8b308d80b47a8d312" target="_top">
|
|
624
|
+number</a> <a class="ulink" href="https://gitweb.torproject.org/tor-browser.git/commit/?h=tor-browser-45.8.0esr-6.5-2&id=98bf6c81b22cb5e4651a5fc060182f27b26c8ee5" target="_top">
|
|
625
|
+of</a> <a class="ulink" href="https://gitweb.torproject.org/tor-browser.git/commit/?h=tor-browser-45.8.0esr-6.5-2&id=14b723f28a6b1dd78093691013d1bf7d49dc4413" target="_top">Mozilla patches</a>.
|
|
626
|
+Further down on our road to proxy safety we <a class="ulink" href="https://gitweb.torproject.org/tor-browser.git/commit/?h=tor-browser-45.8.0esr-6.5-2&id=a9e1d8eac28abb364bbfd3adabeae287751a6a8e" target="_top">
|
|
627
|
+disabled the network tickler</a> as it has the capability to send UDP
|
|
628
|
+traffic.
|
|
629
|
+ </p><p>
|
|
630
|
+
|
|
631
|
+Finally, we <a class="ulink" href="https://gitweb.torproject.org/tor-browser.git/commit/?h=tor-browser-45.8.0esr-6.5-2&id=8e52265653ab223dc5af679f9f0c073b44371fa4" target="_top">
|
|
632
|
+disabled mDNS support</a>, since mDNS uses UDP packets. We also disable
|
|
633
|
+Mozilla's TCPSocket by setting
|
|
634
|
+<span class="command"><strong>dom.mozTCPSocket.enabled</strong></span> to <span class="command"><strong>false</strong></span>. We
|
|
635
|
+<a class="ulink" href="https://trac.torproject.org/projects/tor/ticket/18866" target="_top">intend to
|
|
636
|
+rip out</a> the TCPSocket code in the future to have an even more solid
|
|
637
|
+guarantee that it won't be used by accident.
|
|
638
|
+
|
|
639
|
+ </p><p>
|
584
|
640
|
During every Extended Support Release transition, we perform <a class="ulink" href="https://gitweb.torproject.org/tor-browser-spec.git/tree/audits" target="_top">in-depth
|
585
|
641
|
code audits</a> to verify that there were no system calls or XPCOM
|
586
|
642
|
activity in the source tree that did not use the browser proxy settings.
|
...
|
...
|
@@ -590,12 +646,12 @@ We have verified that these settings and patches properly proxy HTTPS, OCSP,
|
590
|
646
|
HTTP, FTP, gopher (now defunct), DNS, SafeBrowsing Queries, all JavaScript
|
591
|
647
|
activity, including HTML5 audio and video objects, addon updates, WiFi
|
592
|
648
|
geolocation queries, searchbox queries, XPCOM addon HTTPS/HTTP activity,
|
593
|
|
-WebSockets, and live bookmark updates. We have also verified that IPv6
|
594
|
|
-connections are not attempted, through the proxy or otherwise (Tor does not
|
595
|
|
-yet support IPv6). We have also verified that external protocol helpers, such
|
596
|
|
-as SMB URLs and other custom protocol handlers are all blocked.
|
597
|
|
-
|
598
|
|
- </p></li><li class="listitem"><span class="command"><strong>Disabling plugins</strong></span><p>Plugins have the ability to make arbitrary OS system calls and bypass proxy settings. This includes
|
|
649
|
+WebSockets, and live bookmark updates. We have also verified that external
|
|
650
|
+protocol helpers, such as SMB URLs and other custom protocol handlers are all
|
|
651
|
+blocked.
|
|
652
|
+ </p></li><li class="listitem"><span class="command"><strong>Disabling plugins</strong></span><p>
|
|
653
|
+Plugins, like Flash, have the ability to make arbitrary OS system calls and
|
|
654
|
+<a class="ulink" href="http://decloak.net/" target="_top">bypass proxy settings</a>. This includes
|
599
|
655
|
the ability to make UDP sockets and send arbitrary data independent of the
|
600
|
656
|
browser proxy settings.
|
601
|
657
|
</p><p>
|
...
|
...
|
@@ -611,10 +667,28 @@ restricted from automatic load through Firefox's click-to-play preference
|
611
|
667
|
|
612
|
668
|
In addition, to reduce any unproxied activity by arbitrary plugins at load
|
613
|
669
|
time, and to reduce the fingerprintability of the installed plugin list, we
|
614
|
|
-also patch the Firefox source code to <a class="ulink" href="https://gitweb.torproject.org/tor-browser.git/commit/?h=tor-browser-31.6.0esr-4.5-1&id=465cb8295db58a6450dc14a593d29372cbebc71d" target="_top">
|
|
670
|
+also patch the Firefox source code to <a class="ulink" href="https://gitweb.torproject.org/tor-browser.git/commit/?h=tor-browser-45.8.0esr-6.5-2&id=09883246904ce4dede9f3c4d4bb8d644aefe9d1d" target="_top">
|
615
|
671
|
prevent the load of any plugins except for Flash and Gnash</a>. Even for
|
616
|
|
-Flash and Gnash, we also patch Firefox to <a class="ulink" href="https://gitweb.torproject.org/tor-browser.git/commit/?h=tor-browser-31.6.0esr-4.5-1&id=e5531b1baa3c96dee7d8d4274791ff393bafd241" target="_top">prevent loading them into the
|
617
|
|
-address space</a> until they are explicitly enabled.
|
|
672
|
+Flash and Gnash, we also patch Firefox to <a class="ulink" href="https://gitweb.torproject.org/tor-browser.git/commit/?h=tor-browser-45.8.0esr-6.5-2&id=9a0d506e3655f2fdec97ee4217f354941e39b5b3" target="_top">
|
|
673
|
+prevent loading them into the address space</a> until they are explicitly
|
|
674
|
+enabled.
|
|
675
|
+ </p><p>
|
|
676
|
+With <a class="ulink" href="https://wiki.mozilla.org/GeckoMediaPlugins" target="_top">Gecko Media
|
|
677
|
+Plugins</a> (GMPs) a second type of plugins is available. They are mainly
|
|
678
|
+third party codecs and <a class="ulink" href="https://www.w3.org/TR/encrypted-media/" target="_top">EME</a>
|
|
679
|
+content decryption modules. We currently disable these plugins as they either
|
|
680
|
+can't be built reproducibly or are binary blobs which we are not allowed to
|
|
681
|
+audit (or both). For the EME case we use the <span class="command"><strong>--disable-eme</strong></span>
|
|
682
|
+configure switch and set
|
|
683
|
+<span class="command"><strong>browser.eme.ui.enabled</strong></span>,
|
|
684
|
+<span class="command"><strong>media.gmp-eme-adobe.enabled</strong></span>,
|
|
685
|
+<span class="command"><strong>media.eme.enabled</strong></span>, and
|
|
686
|
+<span class="command"><strong>media.eme.apiVisible</strong></span> to <span class="command"><strong>false</strong></span> to indicate
|
|
687
|
+to the user that this feature is disabled. For GMPs in general we make sure that
|
|
688
|
+the external server is not even pinged for updates/downloads in the first place
|
|
689
|
+by setting <span class="command"><strong>media.gmp-manager.url.override</strong></span> to
|
|
690
|
+<span class="command"><strong>data:text/plain,</strong></span> and avoid any UI with <span class="command"><strong>
|
|
691
|
+media.gmp-provider.enabled</strong></span> set to <span class="command"><strong>false</strong></span>.
|
618
|
692
|
|
619
|
693
|
</p></li><li class="listitem"><span class="command"><strong>External App Blocking and Drag Event Filtering</strong></span><p>
|
620
|
694
|
|
...
|
...
|
@@ -642,42 +716,39 @@ bypassing Tor. It is for this reason we disable the addon whitelist
|
642
|
716
|
before installing addons regardless of the source. We also exclude
|
643
|
717
|
system-level addons from the browser through the use of
|
644
|
718
|
<span class="command"><strong>extensions.enabledScopes</strong></span> and
|
645
|
|
-<span class="command"><strong>extensions.autoDisableScopes</strong></span>.
|
|
719
|
+<span class="command"><strong>extensions.autoDisableScopes</strong></span>. Furthermore, we set
|
|
720
|
+<span class="command"><strong>extensions.systemAddon.update.url</strong></span> and <span class="command"><strong>
|
|
721
|
+extensions.hotfix.id</strong></span> to an empty string in order
|
|
722
|
+to avoid the risk of getting extensions installed by Mozilla into Tor Browser.
|
646
|
723
|
|
647
|
724
|
</p></li></ol></div></div><div class="sect2"><div class="titlepage"><div><div><h3 class="title"><a id="state-separation"></a>4.2. State Separation</h3></div></div></div><p>
|
648
|
725
|
|
649
|
726
|
Tor Browser State is separated from existing browser state through use of a
|
650
|
727
|
custom Firefox profile, and by setting the $HOME environment variable to the
|
651
|
|
-root of the bundle's directory. The browser also does not load any
|
|
728
|
+root of the bundle's directory. The browser also does not load any
|
652
|
729
|
system-wide extensions (through the use of
|
653
|
730
|
<span class="command"><strong>extensions.enabledScopes</strong></span> and
|
654
|
731
|
<span class="command"><strong>extensions.autoDisableScopes</strong></span>). Furthermore, plugins are
|
655
|
732
|
disabled, which prevents Flash cookies from leaking from a pre-existing Flash
|
656
|
733
|
directory.
|
657
|
734
|
|
658
|
|
- </p></div><div class="sect2"><div class="titlepage"><div><div><h3 class="title"><a id="disk-avoidance"></a>4.3. Disk Avoidance</h3></div></div></div><div class="sect3"><div class="titlepage"><div><div><h4 class="title"><a id="idp55029872"></a>Design Goal:</h4></div></div></div><div class="blockquote"><blockquote class="blockquote">
|
|
735
|
+ </p></div><div class="sect2"><div class="titlepage"><div><div><h3 class="title"><a id="disk-avoidance"></a>4.3. Disk Avoidance</h3></div></div></div><div class="sect3"><div class="titlepage"><div><div><h4 class="title"><a id="idm357"></a>Design Goal:</h4></div></div></div><div class="blockquote"><blockquote class="blockquote">
|
659
|
736
|
|
660
|
737
|
The User Agent MUST (at user option) prevent all disk records of browser activity.
|
661
|
|
-The user should be able to optionally enable URL history and other history
|
662
|
|
-features if they so desire.
|
663
|
|
-
|
664
|
|
- </blockquote></div></div><div class="sect3"><div class="titlepage"><div><div><h4 class="title"><a id="idp55031232"></a>Implementation Status:</h4></div></div></div><div class="blockquote"><blockquote class="blockquote">
|
665
|
|
-
|
666
|
|
-We achieve this goal through several mechanisms. First, we set the Firefox
|
667
|
|
-Private Browsing preference
|
668
|
|
-<span class="command"><strong>browser.privatebrowsing.autostart</strong></span>. In addition, four Firefox patches are needed to prevent disk writes, even if
|
669
|
|
-Private Browsing Mode is enabled. We need to
|
670
|
|
-
|
671
|
|
-<a class="ulink" href="https://gitweb.torproject.org/tor-browser.git/commit/?h=tor-browser-31.6.0esr-4.5-1&id=44b8ae43a83191bbf5161cbdbf399e10c1b943d0" target="_top">prevent
|
672
|
|
-the permissions manager from recording HTTPS STS state</a>, <a class="ulink" href="https://gitweb.torproject.org/tor-browser.git/commit/?h=tor-browser-31.6.0esr-4.5-1&id=e5abcb28f131aa96e8762212573488d303b3614d" target="_top">prevent
|
673
|
|
-intermediate SSL certificates from being recorded</a>, <a class="ulink" href="https://gitweb.torproject.org/tor-browser.git/commit/?h=tor-browser-31.6.0esr-4.5-1&id=ee34e122ac2929a7668314483e36e58a88c98c08" target="_top">prevent
|
674
|
|
-the clipboard cache from being written to disk for large pastes</a>, and
|
675
|
|
-<a class="ulink" href="https://gitweb.torproject.org/tor-browser.git/commit/?h=tor-browser-31.6.0esr-4.5-1&id=c8e357740dd7bafa2a129007f27d2b243e36f4a2" target="_top">prevent
|
676
|
|
-the content preferences service from recording site zoom</a>. We also had
|
677
|
|
-to disable the media cache with the pref <span class="command"><strong>media.cache_size</strong></span>,
|
678
|
|
-to prevent HTML5 videos from being written to the OS temporary directory,
|
679
|
|
-which happened regardless of the private browsing mode setting.
|
680
|
|
-
|
|
738
|
+The user SHOULD be able to optionally enable URL history and other history
|
|
739
|
+features if they so desire.
|
|
740
|
+
|
|
741
|
+ </blockquote></div></div><div class="sect3"><div class="titlepage"><div><div><h4 class="title"><a id="idm360"></a>Implementation Status:</h4></div></div></div><div class="blockquote"><blockquote class="blockquote">
|
|
742
|
+
|
|
743
|
+ We are working towards this goal through several mechanisms. First, we set
|
|
744
|
+ the Firefox Private Browsing preference
|
|
745
|
+ <span class="command"><strong>browser.privatebrowsing.autostart</strong></span>.
|
|
746
|
+ We also had to disable the media cache with the pref <span class="command"><strong>media.cache_size</strong></span>, to prevent HTML5 videos from being written to the OS temporary directory, which happened regardless of the private browsing mode setting.
|
|
747
|
+ Finally, we needed to disable asm.js as it turns out that
|
|
748
|
+ <a class="ulink" href="https://bugzilla.mozilla.org/show_bug.cgi?id=1047105" target="_top">asm.js
|
|
749
|
+ cache entries get written to disk</a> in private browsing mode. This
|
|
750
|
+ is done by setting <span class="command"><strong>javascript.options.asmjs</strong></span> to
|
|
751
|
+ <span class="command"><strong>false</strong></span> (for linkability concerns with asm.js see below).
|
681
|
752
|
</blockquote></div><div class="blockquote"><blockquote class="blockquote">
|
682
|
753
|
|
683
|
754
|
As an additional defense-in-depth measure, we set the following preferences:
|
...
|
...
|
@@ -699,18 +770,17 @@ auditing work to ensure that yet.
|
699
|
770
|
|
700
|
771
|
For more details on disk leak bugs and enhancements, see the <a class="ulink" href="https://trac.torproject.org/projects/tor/query?keywords=~tbb-disk-leak&status=!closed" target="_top">tbb-disk-leak tag in our bugtracker</a></blockquote></div></div></div><div class="sect2"><div class="titlepage"><div><div><h3 class="title"><a id="app-data-isolation"></a>4.4. Application Data Isolation</h3></div></div></div><p>
|
701
|
772
|
|
702
|
|
-Tor Browser Bundle MUST NOT cause any information to be written outside of the
|
703
|
|
-bundle directory. This is to ensure that the user is able to completely and
|
704
|
|
-safely remove the bundle without leaving other traces of Tor usage on their
|
705
|
|
-computer.
|
|
773
|
+Tor Browser MUST NOT cause any information to be written outside of the bundle
|
|
774
|
+directory. This is to ensure that the user is able to completely and
|
|
775
|
+safely remove it without leaving other traces of Tor usage on their computer.
|
706
|
776
|
|
707
|
777
|
</p><p>
|
708
|
778
|
|
709
|
|
-To ensure TBB directory isolation, we set
|
|
779
|
+To ensure Tor Browser directory isolation, we set
|
710
|
780
|
<span class="command"><strong>browser.download.useDownloadDir</strong></span>,
|
711
|
781
|
<span class="command"><strong>browser.shell.checkDefaultBrowser</strong></span>, and
|
712
|
782
|
<span class="command"><strong>browser.download.manager.addToRecentDocs</strong></span>. We also set the
|
713
|
|
-$HOME environment variable to be the TBB extraction directory.
|
|
783
|
+$HOME environment variable to be the Tor Browser extraction directory.
|
714
|
784
|
</p></div><div class="sect2"><div class="titlepage"><div><div><h3 class="title"><a id="identifier-linkability"></a>4.5. Cross-Origin Identifier Unlinkability</h3></div></div></div><p>
|
715
|
785
|
|
716
|
786
|
The Cross-Origin Identifier Unlinkability design requirement is satisfied
|
...
|
...
|
@@ -733,15 +803,15 @@ the URL bar origin for which browser state exists, possibly with a
|
733
|
803
|
context-menu option to drill down into specific types of state or permissions.
|
734
|
804
|
An example of this simplification can be seen in Figure 1.
|
735
|
805
|
|
736
|
|
- </p><div class="figure"><a id="idp55052928"></a><p class="title"><strong>Figure 1. Improving the Privacy UI</strong></p><div class="figure-contents"><div class="mediaobject" align="center"><img src="NewCookieManager.png" align="middle" alt="Improving the Privacy UI" /></div><div class="caption"><p></p>
|
|
806
|
+ </p><div class="figure"><a id="idm393"></a><p class="title"><strong>Figure 1. Improving the Privacy UI</strong></p><div class="figure-contents"><div class="mediaobject" align="center"><img src="NewCookieManager.png" align="middle" alt="Improving the Privacy UI" /></div><div class="caption"><p></p>
|
737
|
807
|
|
738
|
808
|
This example UI is a mock-up of how isolating identifiers to the URL bar
|
739
|
|
-origin can simplify the privacy UI for all data - not just cookies. Once
|
|
809
|
+domain can simplify the privacy UI for all data - not just cookies. Once
|
740
|
810
|
browser identifiers and site permissions operate on a URL bar basis, the same
|
741
|
811
|
privacy window can represent browsing history, DOM Storage, HTTP Auth, search
|
742
|
812
|
form history, login values, and so on within a context menu for each site.
|
743
|
813
|
|
744
|
|
-</div></div></div><br class="figure-break" /><div class="sect3"><div class="titlepage"><div><div><h4 class="title"><a id="idp55056352"></a>Identifier Unlinkability Defenses in the Tor Browser</h4></div></div></div><p>
|
|
814
|
+</div></div></div><br class="figure-break" /><div class="sect3"><div class="titlepage"><div><div><h4 class="title"><a id="idm400"></a>Identifier Unlinkability Defenses in the Tor Browser</h4></div></div></div><p>
|
745
|
815
|
|
746
|
816
|
Unfortunately, many aspects of browser state can serve as identifier storage,
|
747
|
817
|
and no other browser vendor or standards body has invested the effort to
|
...
|
...
|
@@ -761,45 +831,61 @@ Firefox versions.
|
761
|
831
|
|
762
|
832
|
As a stopgap to satisfy our design requirement of unlinkability, we currently
|
763
|
833
|
entirely disable 3rd party cookies by setting
|
764
|
|
-<span class="command"><strong>network.cookie.cookieBehavior</strong></span> to 1. We would prefer that
|
765
|
|
-third party content continue to function, but we believe the requirement for
|
766
|
|
-unlinkability trumps that desire.
|
|
834
|
+<span class="command"><strong>network.cookie.cookieBehavior</strong></span> to <span class="command"><strong>1</strong></span>. We
|
|
835
|
+would prefer that third party content continue to function, but we believe the
|
|
836
|
+requirement for unlinkability trumps that desire.
|
767
|
837
|
|
768
|
|
- </p></li><li class="listitem"><span class="command"><strong>Cache</strong></span><p>
|
|
838
|
+ </p></li><li class="listitem"><span class="command"><strong>Cache</strong></span><p><span class="command"><strong>Design Goal:</strong></span>
|
|
839
|
+ All cache entries MUST be isolated to the URL bar domain.
|
|
840
|
+ </p><p><span class="command"><strong>Implementation Status:</strong></span>
|
769
|
841
|
|
770
|
|
-In Firefox, there are actually two distinct caching mechanisms: One for
|
771
|
|
-general content (HTML, Javascript, CSS), and one specifically for images. The
|
772
|
|
-content cache is isolated to the URL bar domain by <a class="ulink" href="https://gitweb.torproject.org/tor-browser.git/commit/?h=tor-browser-31.6.0esr-4.5-1&id=7c58be929777d386a03e1faaee648909151fd951" target="_top">altering
|
|
842
|
+In Firefox, there are actually several distinct caching mechanisms: One is for
|
|
843
|
+general content (HTML, JavaScript, CSS). That content cache is isolated to the
|
|
844
|
+URL bar domain by <a class="ulink" href="https://gitweb.torproject.org/tor-browser.git/commit/?h=tor-browser-45.8.0esr-6.5-2&id=9e88ab764b1c9c5d26a398ec6381eef88689929c" target="_top">altering
|
773
|
845
|
each cache key</a> to include an additional ID that includes the URL bar
|
774
|
846
|
domain. This functionality can be observed by navigating to <a class="ulink" href="about:cache" target="_top">about:cache</a> and viewing the key used for each cache
|
775
|
|
-entry. Each third party element should have an additional "id=string"
|
776
|
|
-property prepended, which will list the FQDN that was used to source it.
|
|
847
|
+entry. Each third party element should have an additional "string@:"
|
|
848
|
+property prepended, which will list the base domain that was used to source it.
|
777
|
849
|
|
778
|
850
|
</p><p>
|
779
|
851
|
|
780
|
|
-Additionally, because the image cache is a separate entity from the content
|
781
|
|
-cache, we had to patch Firefox to also <a class="ulink" href="https://gitweb.torproject.org/tor-browser.git/commit/?h=tor-browser-31.6.0esr-4.5-1&id=d8b98a75fb200268c40886d876adc19e00b933bf" target="_top">isolate
|
|
852
|
+Additionally, there is the image cache. Because it is a separate entity from
|
|
853
|
+the content cache, we had to patch Firefox to also <a class="ulink" href="https://gitweb.torproject.org/tor-browser.git/commit/?h=tor-browser-45.8.0esr-6.5-2&id=05749216781d470ab95c2d101dd28ad000d9161f" target="_top">isolate
|
782
|
854
|
this cache per URL bar domain</a>.
|
783
|
855
|
|
|
856
|
+ </p><p>
|
|
857
|
+Furthermore there is the Cache API (CacheStorage). That one is currently not
|
|
858
|
+available in Tor Browser as we do not allow third party cookies and are in
|
|
859
|
+Private Browsing Mode by default.
|
|
860
|
+ </p><p>
|
|
861
|
+Finally, we have the asm.js cache. The cache entry of the sript is (among
|
|
862
|
+others things, like type of CPU, build ID, source characters of the asm.js
|
|
863
|
+module etc.) keyed <a class="ulink" href="https://blog.mozilla.org/luke/2014/01/14/asm-js-aot-compilation-and-startup-performance/" target="_top">to the origin of the script</a>.
|
|
864
|
+Lacking a good solution for binding it to the URL bar domain instead (and given
|
|
865
|
+the storage of asm.js modules in Private Browsing Mode) we decided to disable
|
|
866
|
+asm.js for the time being by setting <span class="command"><strong>javascript.options.asmjs</strong></span> to
|
|
867
|
+<span class="command"><strong>false</strong></span>. It remains to be seen whether keying the cache entry
|
|
868
|
+to the source characters of the asm.js module helps to avoid using it for
|
|
869
|
+cross-origin tracking of users. We did not investigate that yet.
|
784
|
870
|
</p></li><li class="listitem"><span class="command"><strong>HTTP Authentication</strong></span><p>
|
785
|
871
|
|
786
|
872
|
HTTP Authorization headers can be used to encode <a class="ulink" href="http://jeremiahgrossman.blogspot.com/2007/04/tracking-users-without-cookies.html" target="_top">silent
|
787
|
873
|
third party tracking identifiers</a>. To prevent this, we remove HTTP
|
788
|
|
-authentication tokens for third party elements through a <a class="ulink" href="https://gitweb.torproject.org/tor-browser.git/commit/?h=tor-browser-31.6.0esr-4.5-1&id=b8ce4a0760759431f146c71184c89fbd5e1a27e4" target="_top">patch
|
|
874
|
+authentication tokens for third party elements through a <a class="ulink" href="https://gitweb.torproject.org/tor-browser.git/commit/?h=tor-browser-45.8.0esr-6.5-2&id=5e686c690cbc33cf3fdf984e6f3d3fe7b4d83701" target="_top">patch
|
789
|
875
|
to nsHTTPChannel</a>.
|
790
|
876
|
|
791
|
877
|
</p></li><li class="listitem"><span class="command"><strong>DOM Storage</strong></span><p>
|
792
|
878
|
|
793
|
|
-DOM storage for third party domains MUST be isolated to the URL bar origin,
|
|
879
|
+DOM storage for third party domains MUST be isolated to the URL bar domain,
|
794
|
880
|
to prevent linkability between sites. This functionality is provided through a
|
795
|
|
-<a class="ulink" href="https://gitweb.torproject.org/tor-browser.git/commit/?h=tor-browser-31.6.0esr-4.5-1&id=97490c4a90ca1c43374486d9ec0c5593d5fe5720" target="_top">patch
|
|
881
|
+<a class="ulink" href="https://gitweb.torproject.org/tor-browser.git/commit/?h=tor-browser-45.8.0esr-6.5-2&id=20fee895321a7a18e79547e74f6739786558c0e8" target="_top">patch
|
796
|
882
|
to Firefox</a>.
|
797
|
883
|
|
798
|
884
|
</p></li><li class="listitem"><span class="command"><strong>Flash cookies</strong></span><p><span class="command"><strong>Design Goal:</strong></span>
|
799
|
885
|
|
800
|
886
|
Users should be able to click-to-play flash objects from trusted sites. To
|
801
|
|
-make this behavior unlinkable, we wish to include a settings file for all platforms that disables flash
|
802
|
|
-cookies using the <a class="ulink" href="http://www.macromedia.com/support/documentation/en/flashplayer/help/settings_manager03.html" target="_top">Flash
|
|
887
|
+make this behavior unlinkable, we wish to include a settings file for all
|
|
888
|
+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
|
803
|
889
|
settings manager</a>.
|
804
|
890
|
|
805
|
891
|
</p><p><span class="command"><strong>Implementation Status:</strong></span>
|
...
|
...
|
@@ -811,19 +897,19 @@ file on Windows, so Flash remains difficult to enable.
|
811
|
897
|
</p></li><li class="listitem"><span class="command"><strong>SSL+TLS session resumption</strong></span><p><span class="command"><strong>Design Goal:</strong></span>
|
812
|
898
|
|
813
|
899
|
TLS session resumption tickets and SSL Session IDs MUST be limited to the URL
|
814
|
|
-bar origin.
|
|
900
|
+bar domain.
|
815
|
901
|
|
816
|
902
|
</p><p><span class="command"><strong>Implementation Status:</strong></span>
|
817
|
903
|
|
818
|
|
-We currently clear SSL Session IDs upon <a class="link" href="#new-identity" title="4.7. Long-Term Unlinkability via "New Identity" button">New
|
819
|
|
-Identity</a>, we disable TLS Session Tickets via the Firefox Pref
|
820
|
|
-<span class="command"><strong>security.enable_tls_session_tickets</strong></span>. We disable SSL Session
|
821
|
|
-IDs via a <a class="ulink" href="https://gitweb.torproject.org/tor-browser.git/commit/?h=tor-browser-31.6.0esr-4.5-1&id=a01fb747d4b8b24687de538cb6a1304fe27d9d88" target="_top">patch
|
822
|
|
-to Firefox</a>. To compensate for the increased round trip latency from disabling
|
|
904
|
+We disable TLS Session Tickets and SSL Session IDs by
|
|
905
|
+setting <span class="command"><strong>security.ssl.disable_session_identifiers</strong></span> to
|
|
906
|
+<span class="command"><strong>true</strong></span>.
|
|
907
|
+To compensate for the increased round trip latency from disabling
|
823
|
908
|
these performance optimizations, we also enable
|
824
|
909
|
<a class="ulink" href="https://tools.ietf.org/html/draft-bmoeller-tls-falsestart-00" target="_top">TLS
|
825
|
910
|
False Start</a> via the Firefox Pref
|
826
|
911
|
<span class="command"><strong>security.ssl.enable_false_start</strong></span>.
|
|
912
|
+
|
827
|
913
|
</p></li><li class="listitem"><span class="command"><strong>Tor circuit and HTTP connection linkability</strong></span><p>
|
828
|
914
|
|
829
|
915
|
Tor circuits and HTTP connections from a third party in one URL bar origin
|
...
|
...
|
@@ -831,31 +917,31 @@ MUST NOT be reused for that same third party in another URL bar origin.
|
831
|
917
|
|
832
|
918
|
</p><p>
|
833
|
919
|
|
834
|
|
-This isolation functionality is provided by the combination of a <a class="ulink" href="https://gitweb.torproject.org/tor-browser.git/commit/?h=tor-browser-31.6.0esr-4.5-1&id=b3ea705cc35b79a9ba27323cb3e32d5d004ea113" target="_top">Firefox
|
835
|
|
-patch to allow SOCKS usernames and passwords</a>, as well as a Torbutton
|
|
920
|
+This isolation functionality is provided by a Torbutton
|
836
|
921
|
component that <a class="ulink" href="" target="_top">sets
|
837
|
922
|
the SOCKS username and password for each request</a>. The Tor client has
|
838
|
923
|
logic to prevent connections with different SOCKS usernames and passwords from
|
839
|
|
-using the same Tor circuit. Firefox has existing logic to ensure that connections with
|
840
|
|
-SOCKS proxies do not re-use existing HTTP Keep-Alive connections unless the
|
841
|
|
-proxy settings match. We extended this logic to cover SOCKS username and
|
842
|
|
-password authentication, providing us with HTTP Keep-Alive unlinkability.
|
|
924
|
+using the same Tor circuit. Firefox has existing logic to ensure that
|
|
925
|
+connections with SOCKS proxies do not re-use existing HTTP Keep-Alive
|
|
926
|
+connections unless the proxy settings match.
|
|
927
|
+<a class="ulink" href="https://bugzilla.mozilla.org/show_bug.cgi?id=1200802" target="_top">We extended
|
|
928
|
+this logic</a> to cover SOCKS username and password authentication,
|
|
929
|
+providing us with HTTP Keep-Alive unlinkability.
|
843
|
930
|
|
844
|
931
|
</p></li><li class="listitem"><span class="command"><strong>SharedWorkers</strong></span><p>
|
845
|
932
|
|
846
|
933
|
<a class="ulink" href="https://developer.mozilla.org/en-US/docs/Web/API/SharedWorker" target="_top">SharedWorkers</a>
|
847
|
|
-are a special form of Javascript Worker Threads that have a shared scope
|
848
|
|
-between all threads from the same Javascript origin.
|
849
|
|
- </p><p><span class="command"><strong>Design Goal:</strong></span>
|
|
934
|
+are a special form of JavaScript Worker threads that have a shared scope between
|
|
935
|
+all threads from the same Javascript origin.
|
850
|
936
|
|
851
|
|
-SharedWorker scope MUST be isolated to the URL bar domain. A SharedWorker
|
852
|
|
-launched from a third party from one URL bar domain MUST NOT have access to
|
853
|
|
-the objects created by that same third party loaded under another URL bar domain.
|
854
|
|
-
|
855
|
|
- </p><p><span class="command"><strong>Implementation Status:</strong></span>
|
|
937
|
+ </p><p>
|
856
|
938
|
|
857
|
|
-For now, we disable SharedWorkers via the pref
|
858
|
|
-<span class="command"><strong>dom.workers.sharedWorkers.enabled</strong></span>.
|
|
939
|
+The SharedWorker scope MUST be isolated to the URL bar domain. I.e. a
|
|
940
|
+SharedWorker launched from a third party from one URL bar domain MUST NOT have
|
|
941
|
+access to the objects created by that same third party loaded under another URL
|
|
942
|
+bar domain. This functionality is provided by a
|
|
943
|
+<a class="ulink" href="https://gitweb.torproject.org/tor-browser.git/commit/?h=tor-browser-45.8.0esr-6.5-2&id=d17c11445645908086c8d0af84e970e880f586eb" target="_top">
|
|
944
|
+Firefox patch</a>.
|
859
|
945
|
|
860
|
946
|
</p></li><li class="listitem"><span class="command"><strong>blob: URIs (URL.createObjectURL)</strong></span><p>
|
861
|
947
|
|
...
|
...
|
@@ -867,19 +953,33 @@ web. While this UUID value is neither under control of the site nor
|
867
|
953
|
predictable, it can still be used to tag a set of users that are of high
|
868
|
954
|
interest to an adversary.
|
869
|
955
|
|
870
|
|
- </p><p>
|
|
956
|
+ </p><p><span class="command"><strong>Design Goal:</strong></span>
|
871
|
957
|
|
872
|
958
|
URIs created with URL.createObjectURL MUST be limited in scope to the first
|
873
|
|
-party URL bar domain that created them. We provide this isolation in Tor
|
874
|
|
-Browser via a <a class="ulink" href="https://gitweb.torproject.org/tor-browser.git/commit/?h=tor-browser-31.6.0esr-4.5-1&id=0d67ab406bdd3cf095802cb25c081641aa1f0bcc" target="_top">direct
|
875
|
|
-patch to Firefox</a> and disable URL.createObjectURL in the WebWorker
|
876
|
|
-context as a stopgap, due to an edge case with enforcing this isolation in
|
877
|
|
-WebWorkers.
|
|
959
|
+party URL bar domain that created them.
|
|
960
|
+
|
|
961
|
+ </p><p><span class="command"><strong>Implementation Status:</strong></span>
|
|
962
|
+
|
|
963
|
+We provide the isolation in Tor Browser via a <a class="ulink" href="https://gitweb.torproject.org/tor-browser.git/commit/?h=tor-browser-45.8.0esr-6.5-2&id=7eb0b7b7a9c7257140ae5683718e82f3f0884f4f" target="_top">direct
|
|
964
|
+patch to Firefox</a>. However, downloads of PDF files via the download button in the PDF viewer <a class="ulink" href="https://bugs.torproject.org/17933" target="_top">are not isolated yet</a>.
|
878
|
965
|
|
879
|
|
- </p></li><li class="listitem"><span class="command"><strong>SPDY</strong></span><p>
|
|
966
|
+ </p></li><li class="listitem"><span class="command"><strong>SPDY and HTTP/2</strong></span><p><span class="command"><strong>Design Goal:</strong></span>
|
880
|
967
|
|
881
|
|
-Because SPDY can store identifiers, it is disabled through the
|
882
|
|
-Firefox preference <span class="command"><strong>network.http.spdy.enabled</strong></span>.
|
|
968
|
+SPDY and HTTP/2 connections MUST be isolated to the URL bar domain. Furthermore,
|
|
969
|
+all associated means that could be used for cross-domain user tracking (alt-svc
|
|
970
|
+headers come to mind) MUST adhere to this design principle as well.
|
|
971
|
+
|
|
972
|
+ </p><p><span class="command"><strong>Implementation status:</strong></span>
|
|
973
|
+
|
|
974
|
+SPDY and HTTP/2 are currently disabled by setting the
|
|
975
|
+Firefox preferences <span class="command"><strong>network.http.spdy.enabled</strong></span>,
|
|
976
|
+<span class="command"><strong>network.http.spdy.enabled.v2</strong></span>,
|
|
977
|
+<span class="command"><strong>network.http.spdy.enabled.v3</strong></span>,
|
|
978
|
+<span class="command"><strong>network.http.spdy.enabled.v3-1</strong></span>,
|
|
979
|
+<span class="command"><strong>network.http.spdy.enabled.http2</strong></span>,
|
|
980
|
+<span class="command"><strong>network.http.spdy.enabled.http2draft</strong></span>,
|
|
981
|
+<span class="command"><strong>network.http.altsvc.enabled</strong></span>, and
|
|
982
|
+<span class="command"><strong>network.http.altsvc.oe</strong></span> to <span class="command"><strong>false</strong></span>.
|
883
|
983
|
|
884
|
984
|
</p></li><li class="listitem"><span class="command"><strong>Automated cross-origin redirects</strong></span><p><span class="command"><strong>Design Goal:</strong></span>
|
885
|
985
|
|
...
|
...
|
@@ -926,32 +1026,112 @@ We disable the password saving functionality in the browser as part of our
|
926
|
1026
|
since users may decide to re-enable disk history records and password saving,
|
927
|
1027
|
we also set the <a class="ulink" href="http://kb.mozillazine.org/Signon.autofillForms" target="_top">signon.autofillForms</a>
|
928
|
1028
|
preference to false to prevent saved values from immediately populating
|
929
|
|
-fields upon page load. Since Javascript can read these values as soon as they
|
|
1029
|
+fields upon page load. Since JavaScript can read these values as soon as they
|
930
|
1030
|
appear, setting this preference prevents automatic linkability from stored passwords.
|
931
|
1031
|
|
932
|
|
- </p></li><li class="listitem"><span class="command"><strong>HSTS supercookies</strong></span><p>
|
|
1032
|
+ </p></li><li class="listitem"><span class="command"><strong>HSTS and HPKP supercookies</strong></span><p>
|
933
|
1033
|
|
934
|
|
-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
|
|
1034
|
+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</a>
|
|
1035
|
+<a class="ulink" href="http://www.radicalresearch.co.uk/lab/hstssupercookies/" target="_top">
|
935
|
1036
|
supercookies</a>. Since HSTS effectively stores one bit of information per domain
|
936
|
1037
|
name, an adversary in possession of numerous domains can use them to construct
|
937
|
1038
|
cookies based on stored HSTS state.
|
938
|
1039
|
|
|
1040
|
+ </p><p>
|
|
1041
|
+
|
|
1042
|
+HPKP provides <a class="ulink" href="https://zyan.scripts.mit.edu/presentations/toorcon2015.pdf" target="_top">
|
|
1043
|
+a mechanism for user tracking</a> across domains as well. It allows abusing the
|
|
1044
|
+requirement to provide a backup pin and the option to report a pin validation
|
|
1045
|
+failure. In a tracking scenario every user gets a unique SHA-256 value serving
|
|
1046
|
+as backup pin. This value is sent back after (deliberate) pin validation failures
|
|
1047
|
+working in fact as a cookie.
|
|
1048
|
+
|
939
|
1049
|
</p><p><span class="command"><strong>Design Goal:</strong></span>
|
940
|
1050
|
|
941
|
|
-There appears to be three options for us: 1. Disable HSTS entirely, and rely
|
942
|
|
-instead on HTTPS-Everywhere to crawl and ship rules for HSTS sites. 2.
|
943
|
|
-Restrict the number of HSTS-enabled third parties allowed per URL bar origin.
|
944
|
|
-3. Prevent third parties from storing HSTS rules. We have not yet decided upon
|
945
|
|
-the best approach.
|
|
1051
|
+HSTS and HPKP MUST be isolated to the URL bar domain.
|
|
1052
|
+
|
|
1053
|
+ </p><p><span class="command"><strong>Implementation Status:</strong></span>
|
|
1054
|
+
|
|
1055
|
+Currently, HSTS and HPKP state is both cleared by <a class="link" href="#new-identity" title="4.7. Long-Term Unlinkability via "New Identity" button">New Identity</a>,
|
|
1056
|
+but we don't defend against the creation and usage of any of these supercookies
|
|
1057
|
+between <span class="command"><strong>New Identity</strong></span> invocations.
|
|
1058
|
+
|
|
1059
|
+ </p></li><li class="listitem"><span class="command"><strong>Broadcast Channels</strong></span><p>
|
|
1060
|
+
|
|
1061
|
+The BroadcastChannel API allows cross-site communication within the same
|
|
1062
|
+origin. However, to avoid cross-origin linkability broadcast channels MUST
|
|
1063
|
+instead be isolated to the URL bar domain.
|
|
1064
|
+
|
|
1065
|
+ </p><p>
|
|
1066
|
+
|
|
1067
|
+We provide the isolation in Tor Browser via a <a class="ulink" href="https://gitweb.torproject.org/tor-browser.git/commit/?h=tor-browser-45.8.0esr-6.5-2&id=3460a38721810b5b7e785e18f202dde20b3434e8" target="_top">direct
|
|
1068
|
+patch to Firefox</a>. If we lack a window for determining the URL bar
|
|
1069
|
+domain (e.g. in some worker contexts) the use of broadcast channels is disabled.
|
|
1070
|
+
|
|
1071
|
+ </p></li><li class="listitem"><span class="command"><strong>OCSP</strong></span><p>
|
|
1072
|
+
|
|
1073
|
+OCSP requests go to Certfication Authorities (CAs) to check for revoked
|
|
1074
|
+certificates. They are sent once the browser is visiting a website via HTTPS and
|
|
1075
|
+no cached results are available. Thus, to avoid information leaks, e.g. to exit
|
|
1076
|
+relays, OCSP requests MUST go over the same circuit as the HTTPS request causing
|
|
1077
|
+them and MUST therefore be isolated to the URL bar domain. The resulting cache
|
|
1078
|
+entries MUST be bound to the URL bar domain as well. This functionality is
|
|
1079
|
+provided by a
|
|
1080
|
+<a class="ulink" href="https://gitweb.torproject.org/tor-browser.git/commit/?h=tor-browser-45.8.0esr-6.5-2&id=7eb1568275acd4fdf61359c9b1e97c2753e7b2be" target="_top">Firefox patch</a>.
|
|
1081
|
+
|
|
1082
|
+ </p></li><li class="listitem"><span class="command"><strong>Favicons</strong></span><p>
|
|
1083
|
+
|
|
1084
|
+When visiting a website its favicon is fetched via a request originating from
|
|
1085
|
+the browser itself (similar to the OCSP mechanism mentioned in the previous
|
|
1086
|
+section). Those requests MUST be isolated to the URL bar domain. This
|
|
1087
|
+functionality is provided by a
|
|
1088
|
+<a class="ulink" href="https://gitweb.torproject.org/tor-browser.git/commit/?h=tor-browser-45.8.0esr-6.5-2&id=f29f3ff28bbc471ea209d2181770677223c394d1" target="_top">Firefox patch</a>.
|
|
1089
|
+
|
|
1090
|
+ </p></li><li class="listitem"><span class="command"><strong>mediasource: URIs and MediaStreams</strong></span><p>
|
|
1091
|
+
|
|
1092
|
+Much like blob URLs, mediasource: URIs and MediaStreams can be used to tag
|
|
1093
|
+users. Therefore, mediasource: URIs and MediaStreams MUST be isolated to the URL bar domain.
|
|
1094
|
+This functionality is part of a
|
|
1095
|
+<a class="ulink" href="https://gitweb.torproject.org/tor-browser.git/commit/?h=tor-browser-45.8.0esr-6.5-2&id=7eb0b7b7a9c7257140ae5683718e82f3f0884f4f" target="_top">Firefox patch</a>
|
|
1096
|
+
|
|
1097
|
+ </p></li><li class="listitem"><span class="command"><strong>Speculative and prefetched connections</strong></span><p>
|
|
1098
|
+
|
|
1099
|
+Firefox provides the feature to <a class="ulink" href="https://www.igvita.com/2015/08/17/eliminating-roundtrips-with-preconnect/" target="_top">connect speculatively</a> to
|
|
1100
|
+remote hosts if that is either indicated in the HTML file (e.g. by
|
|
1101
|
+<a class="ulink" href="https://w3c.github.io/resource-hints/" target="_top">link
|
|
1102
|
+rel="preconnect" and rel="prefetch"</a>) or otherwise deemed beneficial.
|
|
1103
|
+
|
|
1104
|
+ </p><p>
|
|
1105
|
+
|
|
1106
|
+Firefox does not support rel="prerender", and Mozilla has disabled speculative
|
|
1107
|
+connections and rel="preconnect" usage where a proxy is used (see <a class="ulink" href="https://trac.torproject.org/projects/tor/ticket/18762#comment:3" target="_top"> comment
|
|
1108
|
+3 in bug 18762</a> for further details). Explicit prefetching via the
|
|
1109
|
+rel="prefetch" attribute is still performed, however.
|
|
1110
|
+
|
|
1111
|
+ </p><p><span class="command"><strong>Design Goal:</strong></span>
|
|
1112
|
+
|
|
1113
|
+All pre-loaded links and speculative connections MUST be isolated to the URL
|
|
1114
|
+bar domain, if enabled. This includes isolating both Tor circuit use, as well
|
|
1115
|
+as the caching and associate browser state for the prefetched resource.
|
|
1116
|
+
|
|
1117
|
+ </p><p><span class="command"><strong>Implementation Status:</strong></span>
|
|
1118
|
+
|
|
1119
|
+For automatic speculative connects and rel="preconnect", we leave them
|
|
1120
|
+disabled as per the Mozilla default for proxy settings. However, if enabled,
|
|
1121
|
+speculative connects will be isolated to the proper first party Tor circuit by
|
|
1122
|
+the same mechanism as is used for HTTP Keep-alive. This is true for rel="prefetch"
|
|
1123
|
+requests as well. For rel="preconnect", we isolate them <a class="ulink" href="https://gitweb.torproject.org/tor-browser.git/commit/?h=tor-browser-45.8.0esr-6.5-2&id=9126303651785d02f2df0554f391fffba0b0a00e" target="_top">via
|
|
1124
|
+this patch</a>. This isolation makes both preconnecting and cache warming
|
|
1125
|
+via rel=prefetch ineffective for links to domains other than the current URL
|
|
1126
|
+bar domain. For links to the same domain as the URL bar domain, the full cache
|
|
1127
|
+warming benefit is obtained. As an optimization, any preconnecting to domains
|
|
1128
|
+other than the current URL bar domain can thus be disabled (perhaps with the
|
|
1129
|
+exception of frames), but we do not do this. We allow these requests to
|
|
1130
|
+proceed, but we isolate them.
|
946
|
1131
|
|
947
|
|
- </p><p><span class="command"><strong>Implementation Status:</strong></span> Currently, HSTS state is
|
948
|
|
-cleared by <a class="link" href="#new-identity" title="4.7. Long-Term Unlinkability via "New Identity" button">New Identity</a>, but we don't
|
949
|
|
-defend against the creation of these cookies between <span class="command"><strong>New
|
950
|
|
-Identity</strong></span> invocations.
|
951
|
1132
|
</p></li></ol></div><p>
|
952
|
1133
|
For more details on identifier linkability bugs and enhancements, see the <a class="ulink" href="https://trac.torproject.org/projects/tor/query?keywords=~tbb-linkability&status=!closed" target="_top">tbb-linkability tag in our bugtracker</a>
|
953
|
1134
|
</p></div></div><div class="sect2"><div class="titlepage"><div><div><h3 class="title"><a id="fingerprinting-linkability"></a>4.6. Cross-Origin Fingerprinting Unlinkability</h3></div></div></div><p>
|
954
|
|
-
|
955
|
1135
|
Browser fingerprinting is the act of inspecting browser behaviors and features in
|
956
|
1136
|
an attempt to differentiate and track individual users.
|
957
|
1137
|
</p><p>
|
...
|
...
|
@@ -961,11 +1141,13 @@ vectors. Passive fingerprinting makes use of any information the browser
|
961
|
1141
|
provides automatically to a website without any specific action on the part of
|
962
|
1142
|
the website. Active fingerprinting makes use of any information that can be
|
963
|
1143
|
extracted from the browser by some specific website action, usually involving
|
964
|
|
-Javascript. Some definitions of browser fingerprinting also include
|
|
1144
|
+JavaScript. Some definitions of browser fingerprinting also include
|
965
|
1145
|
supercookies and cookie-like identifier storage, but we deal with those issues
|
966
|
1146
|
separately in the <a class="link" href="#identifier-linkability" title="4.5. Cross-Origin Identifier Unlinkability">preceding section on
|
967
|
1147
|
identifier linkability</a>.
|
|
1148
|
+
|
968
|
1149
|
</p><p>
|
|
1150
|
+
|
969
|
1151
|
For the most part, however, we do not differentiate between passive or active
|
970
|
1152
|
fingerprinting sources, since many active fingerprinting mechanisms are very
|
971
|
1153
|
rapid, and can be obfuscated or disguised as legitimate functionality.
|
...
|
...
|
@@ -1022,7 +1204,7 @@ features behind site permissions, or disable them entirely.
|
1022
|
1204
|
On the other hand, because statistical inference of system performance
|
1023
|
1205
|
requires many iterations to achieve accuracy in the face of noise and
|
1024
|
1206
|
concurrent activity, we are less concerned with this mechanism of extracting
|
1025
|
|
-this information. We also expect that reducing the resolution of Javascript's
|
|
1207
|
+this information. We also expect that reducing the resolution of JavaScript's
|
1026
|
1208
|
time sources will significantly increase the duration of execution required to
|
1027
|
1209
|
extract accurate results, and thus make statistical approaches both
|
1028
|
1210
|
unattractive and highly noticeable due to excessive resource consumption.
|
...
|
...
|
@@ -1046,22 +1228,22 @@ it is important to mention that users themselves theoretically might be
|
1046
|
1228
|
fingerprinted through their behavior while interacting with a website. This
|
1047
|
1229
|
behavior includes e.g. keystrokes, mouse movements, click speed, and writing
|
1048
|
1230
|
style. Basic vectors such as keystroke and mouse usage fingerprinting can be
|
1049
|
|
-mitigated by altering Javascript's notion of time. More advanced issues like
|
|
1231
|
+mitigated by altering JavaScript's notion of time. More advanced issues like
|
1050
|
1232
|
writing style fingerprinting are the domain of <a class="ulink" href="https://github.com/psal/anonymouth/blob/master/README.md" target="_top">other tools</a>.
|
1051
|
1233
|
|
1052
|
1234
|
</p></li><li class="listitem"><span class="command"><strong>Browser Vendor and Version Differences</strong></span><p>
|
1053
|
1235
|
|
1054
|
1236
|
Due to vast differences in feature set and implementation behavior even
|
1055
|
|
-between different versions of the same browser, browser vendor and version
|
1056
|
|
-differences are simply not possible to conceal in any realistic way. It
|
1057
|
|
-is only possible to minimize the differences among different installations of
|
1058
|
|
-the same browser vendor and version. We make no effort to mimic any other
|
1059
|
|
-major browser vendor, and in fact most of our fingerprinting defenses serve to
|
1060
|
|
-differentiate Tor Browser users from normal Firefox users. Because of this,
|
1061
|
|
-any study that lumps browser vendor and version differences into its analysis
|
1062
|
|
-of the fingerprintability of a population is largely useless for evaluating
|
1063
|
|
-either attacks or defenses. Unfortunately, this includes popular large-scale
|
1064
|
|
-studies such as <a class="ulink" href="https://panopticlick.eff.org/" target="_top">Panopticlick</a> and <a class="ulink" href="https://amiunique.org/" target="_top">Am I Unique</a>.
|
|
1237
|
+between different (<a class="ulink" href="https://tsyrklevich.net/2014/10/28/abusing-strict-transport-security/" target="_top">minor</a>)
|
|
1238
|
+versions of the same browser, browser vendor and version differences are simply
|
|
1239
|
+not possible to conceal in any realistic way. It is only possible to minimize
|
|
1240
|
+the differences among different installations of the same browser vendor and
|
|
1241
|
+version. We make no effort to mimic any other major browser vendor, and in fact
|
|
1242
|
+most of our fingerprinting defenses serve to differentiate Tor Browser users
|
|
1243
|
+from normal Firefox users. Because of this, any study that lumps browser vendor
|
|
1244
|
+and version differences into its analysis of the fingerprintability of a
|
|
1245
|
+population is largely useless for evaluating either attacks or defenses.
|
|
1246
|
+Unfortunately, this includes popular large-scale studies such as <a class="ulink" href="https://panopticlick.eff.org/" target="_top">Panopticlick</a> and <a class="ulink" href="https://amiunique.org/" target="_top">Am I Unique</a>.
|
1065
|
1247
|
|
1066
|
1248
|
</p></li></ol></div></div><div class="sect3"><div class="titlepage"><div><div><h4 class="title"><a id="fingerprinting-defenses-general"></a>General Fingerprinting Defenses</h4></div></div></div><p>
|
1067
|
1249
|
|
...
|
...
|
@@ -1077,11 +1259,11 @@ work best with.
|
1077
|
1259
|
|
1078
|
1260
|
</p><div class="orderedlist"><ol class="orderedlist" type="1"><li class="listitem"><span class="command"><strong>Value Spoofing</strong></span><p>
|
1079
|
1261
|
|
1080
|
|
-Value spoofing can be used for simple cases where the browser directly
|
1081
|
|
-provides some aspect of the user's configuration details, devices, hardware,
|
1082
|
|
-or operating system directly to a website. It becomes less useful when the
|
1083
|
|
-fingerprinting method relies on behavior to infer aspects of the hardware or
|
1084
|
|
-operating system, rather than obtain them directly.
|
|
1262
|
+Value spoofing can be used for simple cases where the browser provides some
|
|
1263
|
+aspect of the user's configuration details, devices, hardware, or operating
|
|
1264
|
+system directly to a website. It becomes less useful when the fingerprinting
|
|
1265
|
+method relies on behavior to infer aspects of the hardware or operating system,
|
|
1266
|
+rather than obtain them directly.
|
1085
|
1267
|
|
1086
|
1268
|
</p></li><li class="listitem"><span class="command"><strong>Subsystem Modification or Reimplementation</strong></span><p>
|
1087
|
1269
|
|
...
|
...
|
@@ -1124,7 +1306,7 @@ narrow domain or use case, or when there are alternate ways of accomplishing
|
1124
|
1306
|
the same task, these features and/or certain aspects of their functionality
|
1125
|
1307
|
may be simply removed.
|
1126
|
1308
|
|
1127
|
|
- </p></li></ol></div></div><div class="sect3"><div class="titlepage"><div><div><h4 class="title"><a id="idp55149888"></a>Strategies for Defense: Randomization versus Uniformity</h4></div></div></div><p>
|
|
1309
|
+ </p></li></ol></div></div><div class="sect3"><div class="titlepage"><div><div><h4 class="title"><a id="idm608"></a>Strategies for Defense: Randomization versus Uniformity</h4></div></div></div><p>
|
1128
|
1310
|
|
1129
|
1311
|
When applying a form of defense to a specific fingerprinting vector or source,
|
1130
|
1312
|
there are two general strategies available: either the implementation for all
|
...
|
...
|
@@ -1154,7 +1336,7 @@ need for complicated statistics about the variance of the measured behaviors.
|
1154
|
1336
|
Randomization (especially incomplete randomization) may also provide a false
|
1155
|
1337
|
sense of security. When a fingerprinting attempt makes naive use of randomized
|
1156
|
1338
|
information, a fingerprint will appear unstable, but may not actually be
|
1157
|
|
-sufficiently randomized to impede a dedicated adversary. Sophisticated
|
|
1339
|
+sufficiently randomized to impede a dedicated adversary. Sophisticated
|
1158
|
1340
|
fingerprinting mechanisms may either ignore randomized information, or
|
1159
|
1341
|
incorporate knowledge of the distribution and range of randomized values into
|
1160
|
1342
|
the creation of a more stable fingerprint (by either removing the randomness,
|
...
|
...
|
@@ -1230,10 +1412,10 @@ third party software), as well as their internal functionality.
|
1230
|
1412
|
</p><p><span class="command"><strong>Design Goal:</strong></span>
|
1231
|
1413
|
|
1232
|
1414
|
All plugins that have not been specifically audited or sandboxed MUST be
|
1233
|
|
-disabled. To reduce linkability potential, even sandboxed plugins should not
|
|
1415
|
+disabled. To reduce linkability potential, even sandboxed plugins SHOULD NOT
|
1234
|
1416
|
be allowed to load objects until the user has clicked through a click-to-play
|
1235
|
|
-barrier. Additionally, version information should be reduced or obfuscated
|
1236
|
|
-until the plugin object is loaded. For flash, we wish to <a class="ulink" href="https://trac.torproject.org/projects/tor/ticket/3974" target="_top">provide a
|
|
1417
|
+barrier. Additionally, version information SHOULD be reduced or obfuscated
|
|
1418
|
+until the plugin object is loaded. For Flash, we wish to <a class="ulink" href="https://trac.torproject.org/projects/tor/ticket/3974" target="_top">provide a
|
1237
|
1419
|
settings.sol file</a> to disable Flash cookies, and to restrict P2P
|
1238
|
1420
|
features that are likely to bypass proxy settings. We'd also like to restrict
|
1239
|
1421
|
access to fonts and other system information (such as IP address and MAC
|
...
|
...
|
@@ -1255,8 +1437,8 @@ leaking plugin installation information.
|
1255
|
1437
|
|
1256
|
1438
|
After plugins and plugin-provided information, we believe that the <a class="ulink" href="https://developer.mozilla.org/en-US/docs/HTML/Canvas" target="_top">HTML5
|
1257
|
1439
|
Canvas</a> is the single largest fingerprinting threat browsers face
|
1258
|
|
-today. <a class="ulink" href="http://www.w2spconf.com/2012/papers/w2sp12-final4.pdf" target="_top">Initial
|
1259
|
|
-studies</a> show that the Canvas can provide an easy-access fingerprinting
|
|
1440
|
+today. <a class="ulink" href="https://cseweb.ucsd.edu/~hovav/dist/canvas.pdf" target="_top">
|
|
1441
|
+Studies</a> <a class="ulink" href="https://securehomes.esat.kuleuven.be/~gacar/persistent/the_web_never_forgets.pdf" target="_top">show</a> that the Canvas can provide an easy-access fingerprinting
|
1260
|
1442
|
target: The adversary simply renders WebGL, font, and named color data to a
|
1261
|
1443
|
Canvas element, extracts the image buffer, and computes a hash of that image
|
1262
|
1444
|
data. Subtle differences in the video card, font packs, and even font and
|
...
|
...
|
@@ -1271,10 +1453,12 @@ fingerprinting vectors. If WebGL is normalized through software rendering,
|
1271
|
1453
|
system colors were standardized, and the browser shipped a fixed collection of
|
1272
|
1454
|
fonts (see later points in this list), it might not be necessary to create a
|
1273
|
1455
|
canvas permission. However, until then, to reduce the threat from this vector,
|
1274
|
|
-we have patched Firefox to <a class="ulink" href="https://gitweb.torproject.org/tor-browser.git/commit/?h=tor-browser-31.6.0esr-4.5-1&id=6a169ef0166b268b1a27546a17b3d7470330917d" target="_top">prompt before returning valid image data</a> to the Canvas APIs,
|
1275
|
|
-and for <a class="ulink" href="https://gitweb.torproject.org/tor-browser.git/commit/?h=tor-browser-31.6.0esr-4.5-1&id=7d51acca6383732480b49ccdb5506ad6fb92e651" target="_top">access to isPointInPath and related functions</a>.
|
|
1456
|
+we have patched Firefox to <a class="ulink" href="https://gitweb.torproject.org/tor-browser.git/commit/?h=tor-browser-45.8.0esr-6.5-2&id=526e6d0bc5c68d8c409cbaefc231c71973d949cc" target="_top">prompt before returning valid image data</a> to the Canvas APIs,
|
|
1457
|
+and for access to isPointInPath and related functions. Moreover, we put media
|
|
1458
|
+streams on a canvas behind the site permission in that patch as well.
|
1276
|
1459
|
If the user hasn't previously allowed the site in the URL bar to access Canvas
|
1277
|
|
-image data, pure white image data is returned to the Javascript APIs.
|
|
1460
|
+image data, pure white image data is returned to the JavaScript APIs.
|
|
1461
|
+Extracting canvas image data by third parties is not allowed, though.
|
1278
|
1462
|
|
1279
|
1463
|
</p><p>
|
1280
|
1464
|
</p></li><li class="listitem"><span class="command"><strong>Open TCP Port and Local Network Fingerprinting</strong></span><p>
|
...
|
...
|
@@ -1336,49 +1520,51 @@ it via the pref <span class="command"><strong>dom.gamepad.enabled</strong></span
|
1336
|
1520
|
According to the Panopticlick study, fonts provide the most linkability when
|
1337
|
1521
|
they are provided as an enumerable list in file system order, via either the
|
1338
|
1522
|
Flash or Java plugins. However, it is still possible to use CSS and/or
|
1339
|
|
-Javascript to query for the existence of specific fonts. With a large enough
|
|
1523
|
+JavaScript to query for the existence of specific fonts. With a large enough
|
1340
|
1524
|
pre-built list to query, a large amount of fingerprintable information may
|
1341
|
1525
|
still be available, especially given that additional fonts often end up
|
1342
|
1526
|
installed by third party software and for multilingual support.
|
1343
|
1527
|
|
1344
|
|
- </p><p><span class="command"><strong>Design Goal:</strong></span> The sure-fire way to address font
|
1345
|
|
-linkability is to ship the browser with a font for every language, typeface,
|
1346
|
|
-and style, and to only use those fonts at the exclusion of system fonts. We are
|
1347
|
|
-<a class="ulink" href="https://trac.torproject.org/projects/tor/ticket/13313" target="_top">currently
|
1348
|
|
-investigating</a> this approach, and our current favorite font sets for
|
1349
|
|
-this purpose are the <a class="ulink" href="http://www.droidfonts.com/droidfonts/" target="_top">Droid
|
1350
|
|
-fonts</a>, the <a class="ulink" href="http://hangeul.naver.com/" target="_top">Nanum fonts</a>,
|
1351
|
|
-and <a class="ulink" href="https://fedorahosted.org/lohit/" target="_top">Lohit fonts</a>. The Droid
|
1352
|
|
-font set is fairly complete by itself, but Nanum and Lohit have smaller
|
1353
|
|
-versions of many South Asian languages. When combined in a way that chooses the
|
1354
|
|
-smallest font implementations for each locale, these three font sets provide
|
1355
|
|
-coverage for the all languages used on Wikipedia with more than
|
1356
|
|
-10,000 articles, and several others as well, in approximately 3MB of compressed
|
1357
|
|
-overhead. The <a class="ulink" href="https://www.google.com/get/noto/" target="_top">Noto font
|
1358
|
|
-set</a> is another font set that aims for complete coverage, but is
|
1359
|
|
-considerably larger than the combination of the Droid, Nanum, and Lohit fonts.
|
|
1528
|
+ </p><p><span class="command"><strong>Design Goal:</strong></span>Font-based fingerprinting MUST be rendered ineffective</p><p><span class="command"><strong>Implementation Status:</strong></span>
|
1360
|
1529
|
|
1361
|
|
- </p><p><span class="command"><strong>Implementation Status:</strong></span>
|
|
1530
|
+We <a class="ulink" href="https://trac.torproject.org/projects/tor/ticket/13313" target="_top">investigated
|
|
1531
|
+</a>shipping a predefined set of fonts to all of our users allowing only
|
|
1532
|
+those fonts to be used by websites at the exclusion of system fonts. We are
|
|
1533
|
+currently following this approach, which has been <a class="ulink" href="https://www.bamsoftware.com/papers/fontfp.pdf" target="_top">
|
|
1534
|
+suggested</a> <a class="ulink" href="https://cseweb.ucsd.edu/~hovav/dist/canvas.pdf" target="_top">by
|
|
1535
|
+researchers</a> previously. This defense is available for all three
|
|
1536
|
+supported platforms: Windows, macOS, and Linux, although the implementations
|
|
1537
|
+vary in detail.
|
|
1538
|
+
|
|
1539
|
+ </p><p>
|
1362
|
1540
|
|
1363
|
|
-In the meantime while we investigate shipping our own fonts, we disable
|
1364
|
|
-plugins, which prevents font name enumeration. Additionally, we limit both the
|
1365
|
|
-number of font queries from CSS, as well as the total number of fonts that can
|
1366
|
|
-be used in a document <a class="ulink" href="https://gitweb.torproject.org/tor-browser.git/commit/?h=tor-browser-31.6.0esr-4.5-1&id=e78bc05159a79c1358fa9c64e565af9d98c141ee" target="_top">with
|
1367
|
|
-a Firefox patch</a>. We create two prefs,
|
1368
|
|
-<span class="command"><strong>browser.display.max_font_attempts</strong></span> and
|
1369
|
|
-<span class="command"><strong>browser.display.max_font_count</strong></span> for this purpose. Once these
|
1370
|
|
-limits are reached, the browser behaves as if
|
1371
|
|
-<span class="command"><strong>browser.display.use_document_fonts</strong></span> was set.
|
|
1541
|
+For Windows and macOS we use a preference, <span class="command"><strong>font.system.whitelist</strong></span>,
|
|
1542
|
+to restrict fonts being used to those in the whitelist. This functionality is
|
|
1543
|
+provided <a class="ulink" href="https://gitweb.torproject.org/tor-browser.git/commit/?h=tor-browser-45.8.0esr-6.5-2&id=80d233db514a556d7255034ae057b138527cb2ea" target="_top">by a Firefox patch</a>.
|
|
1544
|
+The whitelist for Windows and macOS contains both a set of
|
|
1545
|
+<a class="ulink" href="https://www.google.com/get/noto" target="_top">Noto fonts</a> which we bundle
|
|
1546
|
+and fonts provided by the operating system. For Linux systems we only bundle
|
|
1547
|
+fonts and <a class="ulink" href="https://gitweb.torproject.org/builders/tor-browser-bundle.git/commit/?id=b88443f6d8af62f763b069eb15e008a46d9b468a" target="_top">
|
|
1548
|
+deploy </a> a <span class="command"><strong>fonts.conf</strong></span> file to restrict the browser to
|
|
1549
|
+use those fonts exclusively. In addition to that we set the <span class="command"><strong>font.name*
|
|
1550
|
+</strong></span> preferences for macOS and Linux to make sure that a given code point
|
|
1551
|
+is always displayed with the same font. This is not guaranteed even if we bundle
|
|
1552
|
+all the fonts Tor Browser uses as it can happen that fonts are loaded in a
|
|
1553
|
+different order on different systems. Setting the above mentioned preferences
|
|
1554
|
+works around this issue by specifying the font to use explicitely.
|
1372
|
1555
|
|
1373
|
1556
|
</p><p>
|
1374
|
1557
|
|
1375
|
|
-To improve rendering, we exempt remote <a class="ulink" href="https://developer.mozilla.org/en-US/docs/CSS/@font-face" target="_top">@font-face
|
1376
|
|
-fonts</a> from these counts, and if a font-family CSS rule lists a remote
|
1377
|
|
-font (in any order), we use that font instead of any of the named local fonts.
|
|
1558
|
+Allowing fonts provided by the operating system for Windows and macOS users is
|
|
1559
|
+currently a compromise between fingerprintability resistance and usability
|
|
1560
|
+concerns. We are still investigating the right balance between them and have
|
|
1561
|
+created a <a class="ulink" href="https://trac.torproject.org/projects/tor/ticket/18097" target="_top">
|
|
1562
|
+ticket in our bug tracker</a> to summarize the current state of our defense
|
|
1563
|
+and future work that remains to be done.
|
1378
|
1564
|
|
1379
|
1565
|
</p></li><li class="listitem"><span class="command"><strong>Monitor, Widget, and OS Desktop Resolution</strong></span><p>
|
1380
|
1566
|
|
1381
|
|
-Both CSS and Javascript have access to a lot of information about the screen
|
|
1567
|
+Both CSS and JavaScript have access to a lot of information about the screen
|
1382
|
1568
|
resolution, usable desktop size, OS widget size, toolbar size, title bar size,
|
1383
|
1569
|
and OS desktop widget sizing information that are not at all relevant to
|
1384
|
1570
|
rendering and serve only to provide information for fingerprinting. Since many
|
...
|
...
|
@@ -1399,20 +1585,24 @@ resolution. In addition, to further reduce resolution-based fingerprinting, we
|
1399
|
1585
|
are <a class="ulink" href="https://trac.torproject.org/projects/tor/ticket/7256" target="_top">investigating
|
1400
|
1586
|
zoom/viewport-based mechanisms</a> that might allow us to always report the
|
1401
|
1587
|
same desktop resolution regardless of the actual size of the content window,
|
1402
|
|
-and simply scale to make up the difference. Until then, the user should also
|
1403
|
|
-be informed that maximizing their windows can lead to fingerprintability under
|
1404
|
|
-this scheme.
|
|
1588
|
+and simply scale to make up the difference. As an alternative to zoom-based
|
|
1589
|
+solutions we are testing a
|
|
1590
|
+<a class="ulink" href="https://trac.torproject.org/projects/tor/ticket/14429" target="_top">different
|
|
1591
|
+approach</a> in our alpha series that tries to round the browser window at
|
|
1592
|
+all times to a multiple 200x100 pixels. Regardless which solution we finally
|
|
1593
|
+pick, until it will be available the user should also be informed that
|
|
1594
|
+maximizing their windows can lead to fingerprintability under the current scheme.
|
1405
|
1595
|
|
1406
|
1596
|
</p><p><span class="command"><strong>Implementation Status:</strong></span>
|
1407
|
1597
|
|
1408
|
|
-We automatically resize new browser windows to a 200x100 pixel multiple using
|
1409
|
|
-a window observer <a class="ulink" href="https://gitweb.torproject.org/torbutton.git/tree/src/chrome/content/torbutton.js#n3361" target="_top">based
|
1410
|
|
-on desktop resolution</a>. To minimize the effect of the long tail of large
|
1411
|
|
-monitor sizes, we also cap the window size at 1000 pixels in each direction.
|
1412
|
|
-Additionally, we patch Firefox to use the client content window size <a class="ulink" href="https://gitweb.torproject.org/tor-browser.git/commit/?h=tor-browser-31.6.0esr-4.5-1&id=bd3b1ed32a9c21fdc92fc35f2ec0a41badc378d5" target="_top">for
|
1413
|
|
-window.screen</a>, and to <a class="ulink" href="https://gitweb.torproject.org/tor-browser.git/commit/?h=tor-browser-31.6.0esr-4.5-1&id=a5648c8d80f396caf294d761cc4a9a76c0b33a9d" target="_top">report
|
1414
|
|
-a window.devicePixelRatio of 1.0</a>. Similarly, we <a class="ulink" href="https://gitweb.torproject.org/tor-browser.git/commit/?h=tor-browser-31.6.0esr-4.5-1&id=3c02858027634ffcfbd97047dfdf170c19ca29ec" target="_top">patch
|
1415
|
|
-DOM events to return content window relative points</a>.
|
|
1598
|
+We automatically resize new browser windows to a 200x100 pixel multiple <a class="ulink" href="https://gitweb.torproject.org/tor-browser.git/commit/?h=tor-browser-45.8.0esr-6.5-2&id=7b3e68bd7172d4f3feac11e74c65b06729a502b2" target="_top">based
|
|
1599
|
+on desktop resolution</a> which is provided by a Firefox patch. To minimize
|
|
1600
|
+the effect of the long tail of large monitor sizes, we also cap the window size
|
|
1601
|
+at 1000 pixels in each direction. In addition to that we set
|
|
1602
|
+<span class="command"><strong>privacy.resistFingerprinting</strong></span>
|
|
1603
|
+to <span class="command"><strong>true</strong></span> to use the client content window size for
|
|
1604
|
+window.screen, and to report a window.devicePixelRatio of 1.0. Similarly,
|
|
1605
|
+we use that preference to return content window relative points for DOM events.
|
1416
|
1606
|
|
1417
|
1607
|
We also force popups to open in new tabs (via
|
1418
|
1608
|
<span class="command"><strong>browser.link.open_newwindow.restriction</strong></span>), to avoid
|
...
|
...
|
@@ -1423,7 +1613,7 @@ maximized windows are detrimental to privacy in this mode.
|
1423
|
1613
|
</p></li><li class="listitem"><span class="command"><strong>Display Media information</strong></span><p>
|
1424
|
1614
|
|
1425
|
1615
|
Beyond simple resolution information, a large amount of so-called "Media"
|
1426
|
|
-information is also exported to content. Even without Javascript, CSS has
|
|
1616
|
+information is also exported to content. Even without JavaScript, CSS has
|
1427
|
1617
|
access to a lot of information about the device orientation, system theme
|
1428
|
1618
|
colors, and other desktop and display features that are not at all relevant to
|
1429
|
1619
|
rendering and also user configurable. Most of this
|
...
|
...
|
@@ -1433,18 +1623,19 @@ user and OS theme defined color values</a> to CSS as well.
|
1433
|
1623
|
|
1434
|
1624
|
</p><p><span class="command"><strong>Design Goal:</strong></span>
|
1435
|
1625
|
|
1436
|
|
-CSS should not be able infer anything that the user has configured about their
|
1437
|
|
-computer. Additionally, it should not be able to infer machine-specific
|
|
1626
|
+A website MUST NOT be able infer anything that the user has configured about
|
|
1627
|
+their computer. Additionally, it SHOULD NOT be able to infer machine-specific
|
1438
|
1628
|
details such as screen orientation or type.
|
1439
|
1629
|
|
1440
|
1630
|
</p><p><span class="command"><strong>Implementation Status:</strong></span>
|
1441
|
1631
|
|
1442
|
|
-We patch
|
1443
|
|
-Firefox to <a class="ulink" href="https://gitweb.torproject.org/tor-browser.git/commit/?h=tor-browser-31.6.0esr-4.5-1&id=cf8956b4460107c5b0053c8fc574e34b0a30ec1e" target="_top">report
|
1444
|
|
-a fixed set of system colors to content window CSS</a>, and <a class="ulink" href="https://gitweb.torproject.org/tor-browser.git/commit/?h=tor-browser-31.6.0esr-4.5-1&id=bbc138486e0489b0d559343fa0522df4ee3b3533" target="_top">prevent
|
1445
|
|
-detection of font smoothing on OSX</a>. We also always
|
1446
|
|
-<a class="ulink" href="https://gitweb.torproject.org/tor-browser.git/commit/?h=tor-browser-31.6.0esr-4.5-1&id=e17d60442ab0db92664ff68d90fe7bf737374912" target="_top">report
|
1447
|
|
-landscape-primary</a> for the screen orientation.
|
|
1632
|
+We set <span class="command"><strong>ui.use_standins_for_native_colors</strong></span> to <span class="command"><strong>true
|
|
1633
|
+</strong></span> and provide a <a class="ulink" href="https://gitweb.torproject.org/tor-browser.git/commit/?h=tor-browser-45.8.0esr-6.5-2&id=c6be9ba561a69250c7d5926d90e0112091453643" target="_top">Firefox patch</a>
|
|
1634
|
+to report a fixed set of system colors to content window CSS, and prevent
|
|
1635
|
+detection of font smoothing on macOS with the help of
|
|
1636
|
+<span class="command"><strong>privacy.resistFingerprinting</strong></span>. We also always
|
|
1637
|
+<a class="ulink" href="https://gitweb.torproject.org/tor-browser.git/commit/?h=tor-browser-45.8.0esr-6.5-2&id=5a159c6bfa310b4339555de389ac16cf8e13b3f5" target="_top">
|
|
1638
|
+report landscape-primary</a> for the <a class="ulink" href="https://w3c.github.io/screen-orientation/" target="_top">screen orientation</a>.
|
1448
|
1639
|
|
1449
|
1640
|
</p></li><li class="listitem"><span class="command"><strong>WebGL</strong></span><p>
|
1450
|
1641
|
|
...
|
...
|
@@ -1459,11 +1650,14 @@ vulnerability surface</a>, we deploy a similar strategy against WebGL as
|
1459
|
1650
|
for plugins. First, WebGL Canvases have click-to-play placeholders (provided
|
1460
|
1651
|
by NoScript), and do not run until authorized by the user. Second, we
|
1461
|
1652
|
obfuscate driver information by setting the Firefox preferences
|
1462
|
|
-<span class="command"><strong>webgl.disable-extensions</strong></span> and
|
1463
|
|
-<span class="command"><strong>webgl.min_capability_mode</strong></span>, which reduce the information
|
1464
|
|
-provided by the following WebGL API calls: <span class="command"><strong>getParameter()</strong></span>,
|
1465
|
|
-<span class="command"><strong>getSupportedExtensions()</strong></span>, and
|
1466
|
|
-<span class="command"><strong>getExtension()</strong></span>.
|
|
1653
|
+<span class="command"><strong>webgl.disable-extensions</strong></span>,
|
|
1654
|
+<span class="command"><strong>webgl.min_capability_mode</strong></span>, and
|
|
1655
|
+<span class="command"><strong>webgl.disable-fail-if-major-performance-caveat</strong></span> which reduce
|
|
1656
|
+the information provided by the following WebGL API calls:
|
|
1657
|
+<span class="command"><strong>getParameter()</strong></span>, <span class="command"><strong>getSupportedExtensions()</strong></span>,
|
|
1658
|
+and <span class="command"><strong>getExtension()</strong></span>. To make the minimal WebGL mode usable we
|
|
1659
|
+additionally <a class="ulink" href="https://gitweb.torproject.org/tor-browser.git/commit/?h=tor-browser-45.8.0esr-6.5-2&id=7b0caa1224c3417754d688344eacc97fbbabf7d5" target="_top">
|
|
1660
|
+normalize its properties with a Firefox patch</a>.
|
1467
|
1661
|
|
1468
|
1662
|
</p><p>
|
1469
|
1663
|
|
...
|
...
|
@@ -1471,7 +1665,60 @@ Another option for WebGL might be to use software-only rendering, using a
|
1471
|
1665
|
library such as <a class="ulink" href="http://www.mesa3d.org/" target="_top">Mesa</a>. The use of
|
1472
|
1666
|
such a library would avoid hardware-specific rendering differences.
|
1473
|
1667
|
|
1474
|
|
- </p></li><li class="listitem"><span class="command"><strong>User Agent and HTTP Headers</strong></span><p><span class="command"><strong>Design Goal:</strong></span>
|
|
1668
|
+ </p></li><li class="listitem"><span class="command"><strong>MediaDevices API</strong></span><p>
|
|
1669
|
+The <a class="ulink" href="https://developer.mozilla.org/en-US/docs/Web/API/MediaDevices" target="_top">
|
|
1670
|
+MediaDevices API</a> provides access to connected media input devices like
|
|
1671
|
+cameras and microphones, as well as screen sharing. In particular, it allows web
|
|
1672
|
+content to easily enumerate those devices with <span class="command"><strong>
|
|
1673
|
+MediaDevices.enumerateDevices()</strong></span>. This relies on WebRTC being compiled
|
|
1674
|
+in which we currently don't do. Nevertheless, we disable this feature for now as
|
|
1675
|
+a defense-in-depth by setting <span class="command"><strong>media.peerconnection.enabled</strong></span> and
|
|
1676
|
+<span class="command"><strong>media.navigator.enabled</strong></span> to <span class="command"><strong>false</strong></span>.
|
|
1677
|
+ </p></li><li class="listitem"><span class="command"><strong>MIME Types</strong></span><p>
|
|
1678
|
+
|
|
1679
|
+Which MIME Types are registered with an operating system depends to a great deal
|
|
1680
|
+on the application software and/or drivers a user chose to install. Web pages
|
|
1681
|
+can not only estimate the amount of MIME types registered by checking
|
|
1682
|
+<span class="command"><strong>navigator.mimetypes.length</strong></span>. Rather, they are even able to
|
|
1683
|
+test whether particular MIME types are available which can have a non-negligible
|
|
1684
|
+impact on a user's fingerprint. We prevent both of these information leaks with
|
|
1685
|
+a direct <a class="ulink" href="https://gitweb.torproject.org/tor-browser.git/commit/?h=tor-browser-45.8.0esr-6.5-2&id=38999857761196b0b7f59f49ee93ae13f73c6149" target="_top">Firefox patch</a>.
|
|
1686
|
+
|
|
1687
|
+ </p></li><li class="listitem"><span class="command"><strong>System Uptime</strong></span><p>
|
|
1688
|
+
|
|
1689
|
+It is possible to get the system uptime of a Tor Browser user by querying the
|
|
1690
|
+<span class="command"><strong>Event.timestamp</strong></span> property. We avoid this by setting <span class="command"><strong>
|
|
1691
|
+dom.event.highrestimestamp.enabled</strong></span> to <span class="command"><strong>true</strong></span>.
|
|
1692
|
+
|
|
1693
|
+ </p></li><li class="listitem"><span class="command"><strong>Keyboard Layout Fingerprinting</strong></span><p>
|
|
1694
|
+
|
|
1695
|
+<span class="command"><strong>KeyboardEvent</strong></span>s provide a way for a website to find out
|
|
1696
|
+information about the keyboard layout of its visitors. In fact there are <a class="ulink" href="https://developers.google.com/web/updates/2016/04/keyboardevent-keys-codes" target="_top">
|
|
1697
|
+several dimensions</a> to this fingerprinting vector. The <span class="command"><strong>
|
|
1698
|
+KeyboardEvent.code</strong></span> property represents a physical key that can't be
|
|
1699
|
+changed by the keyboard layout nor by the modifier state. On the other hand the
|
|
1700
|
+<span class="command"><strong>KeyboardEvent.key</strong></span> property contains the character that is
|
|
1701
|
+generated by that key. This is dependent on things like keyboard layout, locale
|
|
1702
|
+and modifier keys.
|
|
1703
|
+
|
|
1704
|
+ </p><p><span class="command"><strong>Design Goal:</strong></span>
|
|
1705
|
+
|
|
1706
|
+Websites MUST NOT be able to infer any information about the keyboard of a Tor
|
|
1707
|
+Browser user.
|
|
1708
|
+
|
|
1709
|
+ </p><p><span class="command"><strong>Implementation Status:</strong></span>
|
|
1710
|
+
|
|
1711
|
+We provide <a class="ulink" href="https://gitweb.torproject.org/tor-browser.git/commit/?h=tor-browser-45.8.0esr-6.5-2&id=a65b5269ff04e4fbbb3689e2adf853543804ffbf" target="_top">two</a>
|
|
1712
|
+<a class="ulink" href="https://gitweb.torproject.org/tor-browser.git/commit/?h=tor-browser-45.8.0esr-6.5-2&id=383b8e7e073ea79e70f19858efe1c5fde64b99cf" target="_top">Firefox patches</a> that
|
|
1713
|
+take care of spoofing <span class="command"><strong>KeyboardEvent.code</strong></span> and <span class="command"><strong>
|
|
1714
|
+KeyboardEvent.keyCode</strong></span> by providing consensus (US-English-style) fake
|
|
1715
|
+properties. This is achieved by hiding the user's use of the numpad, and any
|
|
1716
|
+non-QWERTY US English keyboard. Characters from non-en-US languages
|
|
1717
|
+are currently returning an empty <span class="command"><strong>KeyboardEvent.code</strong></span> and a
|
|
1718
|
+<span class="comman |