root / dotorg / trunk / html / beps / bep_0008.html

Revision 11050, 28.6 kB (checked in by dave, 4 months ago)

rebuild html.

Line 
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">
3<html xmlns="http://www.w3.org/1999/xhtml" xml:lang="en" lang="en">
4<head>
5<meta http-equiv="Content-Type" content="text/html; charset=utf-8" />
6<meta name="generator" content="Docutils 0.5: http://docutils.sourceforge.net/" />
7<title></title>
8<link rel="stylesheet" href="../css/bep.css" type="text/css" />
9</head>
10<body>
11<div class="document">
12
13<div id="upper" class="clear">
14<div id="wrap">
15<div id="header">
16<h1><a href="../index.html">BitTorrent<span>.org</span></a></h1>
17</div>
18<div id="nav">
19<ul>
20<li><a href="../index.html">Home</a></li>
21<li><a href="../introduction.html">For Users</a></li>
22<li><a href="bep_0000.html"><span>For Developers</span></a></li>
23<!-- <li><a href="./blog">Blog</a></li> -->
24<li><a href="http://forum.bittorrent.org"> Forums </li>
25<li><a href="../donate.html">Donate!</a></li>
26</ul>
27</div> <!-- nav -->
28<!-- ### Begin Content ### -->
29<div id="second">
30
31
32
33<table class="rfc2822 docutils field-list" frame="void" rules="none">
34<col class="field-name" />
35<col class="field-body" />
36<tbody valign="top">
37<tr class="field"><th class="field-name">BEP:</th><td class="field-body">8</td>
38</tr>
39<tr class="field"><th class="field-name">Title:</th><td class="field-body">Tracker Peer Obfuscation</td>
40</tr>
41<tr class="field"><th class="field-name">Version:</th><td class="field-body">11049</td>
42</tr>
43<tr class="field"><th class="field-name">Last-Modified:</th><td class="field-body"><a class="reference external" href="http://bittorrent.org/trac/browser/dotorg/trunk/html/beps/bep_0008.rst">2008-04-29 18:09:14 -0700 (Tue, 29 Apr 2008)</a></td>
44</tr>
45<tr class="field"><th class="field-name">Author:</th><td class="field-body">David Harrison &lt;dave&#32;&#97;t&#32;bittorrent.com&gt;, Anthony Ciani &lt;tony&#32;&#97;t&#32;ciani.phy.uic.edu&gt;, Arvid Norberg &lt;arvid&#32;&#97;t&#32;bittorrent.com&gt;, Greg Hazel &lt;greg&#32;&#97;t&#32;bittorrent.com&gt;</td>
46</tr>
47<tr class="field"><th class="field-name">Status:</th><td class="field-body">Deferred</td>
48</tr>
49<tr class="field"><th class="field-name">Type:</th><td class="field-body">Standards Track</td>
50</tr>
51<tr class="field"><th class="field-name">Created:</th><td class="field-body">31-Jan-2008</td>
52</tr>
53<tr class="field"><th class="field-name">Post-History:</th><td class="field-body"></td>
54</tr>
55</tbody>
56</table>
57<hr />
58<div class="contents topic" id="contents">
59<p class="topic-title first">Contents</p>
60<ul class="simple">
61<li><a class="reference internal" href="#announce" id="id8">Announce</a></li>
62<li><a class="reference internal" href="#announce-response" id="id9">Announce Response</a></li>
63<li><a class="reference internal" href="#obfuscation-method" id="id10">Obfuscation Method</a></li>
64<li><a class="reference internal" href="#optimizations" id="id11">Optimizations</a></li>
65<li><a class="reference internal" href="#backwards-compatibility" id="id12">Backwards Compatibility</a></li>
66<li><a class="reference internal" href="#rationale" id="id13">Rationale</a></li>
67<li><a class="reference internal" href="#references" id="id14">References</a></li>
68<li><a class="reference internal" href="#example-python-code" id="id15">Example Python Code</a></li>
69<li><a class="reference internal" href="#copyright" id="id16">Copyright</a></li>
70</ul>
71</div>
72<p>This extends the tracker protocol to support simple obfuscation of the
73peers it returns, using the infohash as a shared secret between the
74peer and the tracker. The obfuscation does not provide any security
75against eavesdroppers that know the infohash of the torrent.  The goal
76is to prevent internet service providers and other network
77administrators from blocking or disrupting bittorrent traffic
78connections that span between the receiver of a tracker response and
79any peer IP-port appearing in that tracker response.</p>
80<p>The key words &quot;MUST&quot;, &quot;MUST NOT&quot;, &quot;REQUIRED&quot;, &quot;SHALL&quot;, &quot;SHALL NOT&quot;, &quot;SHOULD&quot;,
81&quot;SHOULD NOT&quot;, &quot;RECOMMENDED&quot;, &quot;MAY&quot;, and &quot;OPTIONAL&quot; in this document are
82to be interpreted as described in IETF <a class="reference external" href="http://tools.ietf.org/html/rfc2119">RFC 2119</a> <a class="footnote-reference" href="#id6" id="id7">[5]</a>.</p>
83<div class="section" id="announce">
84<h1>Announce</h1>
85<p>When using this extension, instead of passing the <tt class="docutils literal"><span class="pre">info_hash</span></tt> parameter
86to the tracker, a <tt class="docutils literal"><span class="pre">sha_ih</span></tt> is passed.</p>
87<p>The value of <tt class="docutils literal"><span class="pre">sha_ih</span></tt> MUST be the info-hash of the torrent, with a second
88SHA-1 applied to it.</p>
89<p>For example if a torrent has infohash with hex representation
90<tt class="docutils literal"><span class="pre">aaf4c61ddcc5e8a2dabedef3b482cd9aea9434d</span></tt> then its <tt class="docutils literal"><span class="pre">sha_ih</span></tt> is
91<tt class="docutils literal"><span class="pre">sha1(infohash)='6b4f89a54e2d27ecd7e8da5b4ab8fd9d1d8b119'</span></tt>.</p>
92<p>The value MUST be url encoded, just like the <tt class="docutils literal"><span class="pre">info_hash</span></tt>.  Thus the
93<tt class="docutils literal"><span class="pre">sha_ih</span></tt> above when url encoded becomes
94<tt class="docutils literal"><span class="pre">kO%89%A5N-%27%EC%D7%E8%DA%05%B4%AB%8F%D9%D1%D8%B1%19</span></tt>.</p>
95<p>If the <tt class="docutils literal"><span class="pre">sha_ih</span></tt> is passed then the value for the <tt class="docutils literal"><span class="pre">port</span></tt> parameter
96should be treated as a 16 bit integer and MUST be obscured as
97described in the <a class="reference internal" href="#obfuscation-method">Obfuscation Method</a> section.  Similarly if the
98optional <tt class="docutils literal"><span class="pre">ip</span></tt> parameter is passed in the announce then its value
99MUST also be so obscured.</p>
100<p>This extension does not change the semantics of any parameter passed
101in the peer's announce.</p>
102</div>
103<div class="section" id="announce-response">
104<h1>Announce Response</h1>
105<p>If the tracker supports this extension, the response should be exactly
106the same as if the <tt class="docutils literal"><span class="pre">info_hash</span></tt> had been passed, except that any
107field that contains peer information (such as <tt class="docutils literal"><span class="pre">peers</span></tt>, <tt class="docutils literal"><span class="pre">peers6</span></tt> or
108any other field defined by another extension) MUST be obfuscated as
109described in the next section.</p>
110<p>There are additional parameters the tracker may OPTIONALLY return.
111These are discussed in the <a class="reference internal" href="#optimizations">optimizations</a> section.</p>
112</div>
113<div class="section" id="obfuscation-method">
114<h1>Obfuscation Method</h1>
115<p>The values for the <tt class="docutils literal"><span class="pre">ip</span></tt> and <tt class="docutils literal"><span class="pre">port</span></tt> announce parameters, the
116<em>returned peer list</em> and any other values that contain peer
117information are obscured using the method described in this section.</p>
118<p>We distinguish between the <em>tracker peer list</em> and the <em>returned peer
119list</em>.  The <em>tracker peer list</em> contains the ip-port pairs of all
120known peers in a given torrent, i.e., those peers that have reported
121to the tracker that they are transferring the file with a given
122infohash.  The tracker may store this peer list however it wishes.
123The <em>returned peer list</em> contains a packed array of ip-port pairs
124conforming to the BitTorrent protocol specification.  If the swarm is
125sufficiently large then the returned ip-port pairs constitute a subset
126of the ip-port pairs in the <em>tracker peer list</em>.</p>
127<p>When a parameter is obscured, it is encrypted using RC4-drop768
128encryption using the infohash as a shared secret and optionally
129employing an initialization vector.</p>
130<p>For the remainder of this document RC4 refers to RC4-drop768.  In the
131process of encryption, RC4 generates a pseudorandom string that is
132XOR'd with the plaintext to generate the ciphertext.  The receiver
133recovers the plaintext by generating the same pseudorandom string and
134XOR'ing it with the ciphertext.  In generating the pseudorandom
135string, the tracker and client MUST discard the first 768 bytes.  The
136next 8 bytes in the pseudorandom string are reserved for optimizations
137discussed in the next section.</p>
138<p>To communicate an initialization vector, the tracker includes in the
139bencoded response the parameter <tt class="docutils literal"><span class="pre">iv</span></tt> with value set to a byte string
140containing the initialization vector.  The initialization vector can
141be of arbitrary length and is sent in plaintext.  Initialization
142vectors can only be applied to parameters in tracker responses and NOT
143to announces.</p>
144<p>If the tracker sends no initialization vector then the infohash is
145used as the RC4 key (160 bit key).  If the tracker provides an
146initialization vector then the RC4 key is generated by appending the
147vector to the infohash and then hashing with SHA-1.  The resulting
148hash is then used as the RC4 key.</p>
149<p>For example, given infohash <tt class="docutils literal"><span class="pre">aaf4c61ddcc5e8a2dabedef3b482cd9aea9434d</span></tt>
150and initialization vector <tt class="docutils literal"><span class="pre">abcd</span></tt> both represented in hex, the RC4 key
151is derived as follows:</p>
152<pre class="literal-block">
153key = sha1( 'aaf4c61ddcc5e8a2dabedef3b482cd9aea9434dabcd' )
154</pre>
155<p>The resulting key in hex is <tt class="docutils literal"><span class="pre">f36e9cae87cf33e07645ef5ca745a8a83469f31e</span></tt>.</p>
156<p>It is RECOMMENDED that the tracker use the initialization vector, and
157that it change the <tt class="docutils literal"><span class="pre">iv</span></tt> on roughly the same period as the rerequest
158interval.  The reasoning for this is contained in the rationale.</p>
159</div>
160<div class="section" id="optimizations">
161<h1>Optimizations</h1>
162<p>The described optimizations are OPTIONAL for the tracker, but the
163corresponding client-side MUST be implemented by clients that support
164this extension.  These optimizations hobble the strength of the RC4
165encryption in order to improve tracker performance.  In the <a class="reference internal" href="#rationale">rationale</a>
166section we discuss why hobbling RC4 is reasonable and in many cases
167has negligible foreseen effect on security.</p>
168<p>For the purpose of these optimizations we assume that the tracker
169stores the tracker peer list for each infohash as a packed array that
170can be copied directly into the response.  We further assume that the
171packed array is reused many times and that with each request the
172tracker either returns the entire packed array or copies a single
173contiguous substring from the tracker peer list into the response.</p>
174<p>If the peerlist is represented and used as assumed then to improve
175randomness in the set of peers handed out by the tracker, it is
176RECOMMENDED that the tracker periodically reshuffle the peerlist with
177period similar to the rerequest interval.  After each reshuffle the
178tracker reperforms the operations described in this section.</p>
179<p>To reduce computation the tracker MAY cache the pseudorandom string
180generated by RC4 and reuse it as peers arrive and depart.</p>
181<p>The tracker MAY also cache the encrypted tracker peer list.  To
182support this the tracker MUST pass two additional parameters <em>i</em> and <em>n</em>
183each with 32-bit integer values, except the tracker MAY omit <em>i</em> and
184<em>n</em> when <em>i=0</em> and the <em>returned peer list</em> is the entire <em>tracker peer
185list</em>.  Whether the tracker returns <em>i</em> and <em>n</em>, the first 8 bytes of
186the RC4 psuedorandom string are reserved for obscuring <em>i</em> and <em>n</em>.
187We come back to this momentarily.  Decryption starts by XORing from
188<em>6i</em> bytes for ipv4 (or <em>18i</em> for ipv6) into the pseudorandom string
189after the discarded and reserved bytes.  Assuming that the tracker
190encrypted the tracker peer list starting from the first byte after the
191discarded and reserved bytes in the pseudorandom string then <em>i</em> also
192corresponds to the <em>ith</em> ip-port pair in the tracker peer list.</p>
193<p>So that the client and the tracker do not have to generate an
194arbitrarily long pseudorandom string to support large swarms, we
195assume the tracker bounds the length of the pseudorandom string and
196reports the length in ip-port pairs as the value to parameter <em>n</em><em>n</em>
197excludes reserved and discarded bytes.  We RECOMMEND that <em>n</em> be equal
198to the length of the tracker peer list or random but within constant
199factor of the longest peerlist returned by the tracker, whichever is
200smaller.  Thus the tracker encrypts the <em>jth</em> byte of the <em>ith</em>
201ip-port pair in an ipv4 tracker peer list by XORing with the byte
202<em>(6i+j)</em> <cite>mod</cite> <em>n</em> bytes into the pseudorandom string.</p>
203<p>Transmitting <em>i</em> and <em>n</em> as plaintext would significantly reduce the
204cost for an attacker to recover the pseudorandom string.  The tracker
205MUST XOR the value of <em>i</em> with the first 32 bits of the pseudorandom
206string.  The tracker then XORs <em>n</em> with the next 32 bits from the
207pseudorandom string (see Figure 1).</p>
208<div class="figure">
209<img alt="bep_0008_pseudo.png" src="bep_0008_pseudo.png" />
210<p class="caption"><strong>Figure 1:</strong> The first 768 bytes of the RC4 pseudorandom
211string are discarded.  The parameter <em>i</em> in the tracker response has
212value <tt class="docutils literal"><span class="pre">x</span> <span class="pre">xor</span> <span class="pre">i</span></tt>.  The parameter <em>n</em> has value <tt class="docutils literal"><span class="pre">y</span> <span class="pre">xor</span> <span class="pre">n</span></tt>.</p>
213</div>
214<p>We describe encryption in the following example for an ipv4 tracker peer
215list consisting of 3 ip-port pairs, and using an RC4 pseudorandom string
216of length <em>n=2</em>. <em>n</em> is small for purposes of illustration.  Also, for the
217purpose of illustration, the tracker returns only 2 peers at a time.</p>
218<pre class="literal-block">
219Given the following peer list
220(208.72.193.86, 6881), (209.81.173.15,14321), (128.213.6.8, 6881)
221
222As a packed array represented in hex it becomes
223
224d048c1561ae1d151ad0f37f180d506081ae1
225
226which we XOR with an RC4 pseudorandom string excluding discarded and
227reserved bytes, e.g.,
228
229a496e5f9b83e835013d42226
230
231to generate
232
23374de24afa2df5201bedb15d72443e3f1a2df
234</pre>
235<p>Because the RC4 pseudorandom string is shorter than the tracker
236peer list, we wrap to the beginning of the pseudorandom string.</p>
237<p>A tracker returning the first two peers would return the bencoded
238equivalent of:</p>
239<pre class="literal-block">
240peers=74de24afa2df5201bedb15d7, i=0, n=2
241</pre>
242<p>A tracker returning the second and third peer would return the
243bencoded equivalent of:</p>
244<pre class="literal-block">
245peers=5201bedb15d72443e3f1a2df, i=1, n=2
246</pre>
247<p>In each response the tracker includes additional parameters such as
248the rerequest <tt class="docutils literal"><span class="pre">interval</span></tt> and the initialization vector <tt class="docutils literal"><span class="pre">iv</span></tt>.</p>
249<p>The tracker response MUST remain a valid bencoded message.</p>
250</div>
251<div class="section" id="backwards-compatibility">
252<h1>Backwards Compatibility</h1>
253<p>Trackers that support obfuscation are identified in the .torrent file
254by the inclusion of an <tt class="docutils literal"><span class="pre">obfuscate-announce-list</span></tt> which otherwise has the
255same semantics as the <tt class="docutils literal"><span class="pre">announce-list</span></tt> parameter.  Peers that do not support
256obfuscation simply ignore the <tt class="docutils literal"><span class="pre">obfuscate-announce-list</span></tt>.</p>
257<p>A client that is configured to use this extension should always send
258the <tt class="docutils literal"><span class="pre">sha_ih</span></tt> to any tracker supporting obfuscation.  The client
259SHOULD only contact trackers in the <tt class="docutils literal"><span class="pre">announce-list</span></tt> once the client
260has attempted all trackers in the <tt class="docutils literal"><span class="pre">obfuscate-announce-list</span></tt> and all failed.</p>
261<p>If a tracker that supports obfuscation wishes to allow legacy peers to
262connect to the tracker then the announce URL should appear in both the
263<tt class="docutils literal"><span class="pre">obfuscate-announce-list</span></tt> and the <tt class="docutils literal"><span class="pre">announce-list</span></tt>.</p>
264<p>If a tracker URL appears in both lists running on the same port, and
265the tracker failed to respond when selected from the
266<tt class="docutils literal"><span class="pre">obfuscate-announce-list</span></tt> then the client MAY treat the tracker in
267the <tt class="docutils literal"><span class="pre">announce-list</span></tt> as if it were temporarily unreachable and defer
268trying it until it has tried other trackers in the <tt class="docutils literal"><span class="pre">announce-list</span></tt>.</p>
269<p>Peers MUST never send both the <tt class="docutils literal"><span class="pre">info_hash</span></tt> and <tt class="docutils literal"><span class="pre">sha_ih</span></tt> parameters
270in the same request, since that would defeat the purpose of the shared
271secret.</p>
272<p>Any peer that requests with a <tt class="docutils literal"><span class="pre">sha_ih</span></tt> SHOULD implement Message
273Stream Encryption (MSE) <a class="footnote-reference" href="#mse" id="id1">[1]</a>.  Any peer returned from the tracker
274in response to a request with a <tt class="docutils literal"><span class="pre">sha_ih</span></tt> SHOULD be assumed to
275support Message Stream Encryption.  We include these provisions
276because if a peer communicates with another peer without using MSE
277then the BitTorrent protocol is trivially identified from the first
278twenty bytes of the BitTorrent header and the <tt class="docutils literal"><span class="pre">info_hash</span></tt> appears in
279plaintext as the next twenty bytes, hence also defeating the purpose
280of the shared secret.</p>
281<p>If the tracker does not know enough peers assumed to support MSE to
282return the desired number of peers then it MAY include peers that are
283not assumed to support MSE.  If a peer closes a connection in response
284to an encrypted header then the initiating peer SHOULD assume that the
285peer does not support MSE.  The initiating peer however SHOULD ONLY
286initiate unencrypted connections when all peers have been tried and
287those that support MSE fail to provide &quot;adequate performance.&quot;  We
288intentionally omit any definition of &quot;adequate performance.&quot;</p>
289</div>
290<div class="section" id="rationale">
291<h1>Rationale</h1>
292<p>This extension directly addresses a known attack on the BitTorrent
293protocol performed by some deployed network hardware.  By obscuring
294the ip-port pairs network hardware can no longer easily identify
295ip-port pairs that are running BitTorrent by observing peer-to-tracker
296communications.  This deployed hardware under some conditions disrupts
297BitTorrent connections by injecting forged TCP reset packets.</p>
298<p>This hardware was presumably deployed to get around BitTorrent
299Message Stream Encryption <a class="footnote-reference" href="#mse" id="id2">[1]</a>.  Peers implementing BitTorrent Message Stream
300Encryption obfuscate peer-to-peer connections by employing RC4
301encryption on every byte from the first byte transferred. BitTorrent
302Message Stream Encryption thus increases the difficulty for a device
303observing passing packets to identify BitTorrent peer-to-peer
304connections.</p>
305<p>By using the SHA-1 of the infohash, the tracker is able to identify
306torrents without sending the plaintext infohash and without requiring
307an additional prior exchange of a shared secret.  Where trackers now
308maintain mappings from infohash to the corresponding torrent's
309peerlist and other torrent-specific state, obfuscated trackers would
310need one additional mapping from <tt class="docutils literal"><span class="pre">sha_ih</span></tt> to the torrent's state.
311Trackers may also cache the encrypted version of each torrent's
312tracker peer list, to increase computational performance at the
313expense of increasing memory footprint by a constant factor.</p>
314<p>The obfuscation method meets the following criteria:</p>
315<ul class="simple">
316<li>The entire plaintext of the peer list is not easily obtained even if
317an eavesdropper identifies one or more subsequent connections as
318using BitTorrent and the corresponding ip-port pairs appeared in the
319ciphertext of the tracker response.</li>
320<li>Even when a subsequent connection from a peer that has received a
321tracker response is observed by an eavesdropper, it is difficult to
322map the ip-port pair to specific ciphertext to verify that the
323connection is using BitTorrent.</li>
324</ul>
325<p>When the <a class="reference internal" href="#optimizations">optimizations</a> are used,</p>
326<ul class="simple">
327<li>Few computations are performed at request time.</li>
328<li>Encryption may be performed at the time a peer is added.
329The encrypted peer ip and port may be handed out hundreds of times.</li>
330<li>Security is minimally impacted.</li>
331</ul>
332<p>The objective is NOT to create a cryptographically secure protocol
333that can survive unlimited observation of passing packets and
334substantial computational resources on network timescales.  The objective
335is to raise the bar sufficiently to deter attacks based on observing
336ip-port numbers in peer-to-tracker communications.</p>
337<p>If a tracker observes a large number of tracker requests and responses
338and subsequent connections, it is possible to attack the encryption.
339RC4 is known to have a number of weaknesses especially in the way it
340is used with WEP <a class="footnote-reference" href="#borisov" id="id3">[2]</a> <a class="footnote-reference" href="#scott" id="id4">[3]</a> <a class="footnote-reference" href="#stubblefeld" id="id5">[4]</a>.  However, with
341tracker peer obfuscation, the number of bytes transferred between the
342tracker and a client is likely significantly smaller than transferred
343between a wireless computer and a basestation.  An attacker faces a
344much larger task in obtaining sufficient ciphertext to directly break
345the encryption.</p>
346<p>Hobbling the RC4 encryption by using a bounded-length RC4 pseudorandom
347string for small swarms is likely to have negilgible impact on
348security over any other encyption method since the pseudorandom string
349is probably equal to or longer than the plaintext and thus no part of
350it is repeated in the XOR except as peers arrive or leave the swarm.
351Thus on the timescales of rerequest intervals, nearly the same
352ciphertext is handed to every peer requesting the same infohash.
353Intercepting the same ciphertext multiple times provides no additional
354information to the attacker.  The attacker could correlate ip-port
355pairs in connections following tracker responses, but an attacker
356could do this regardless of the encryption method employed.
357Furthermore more direct methods of traffic analysis applied to
358peer-to-peer communication is available to network operators.</p>
359<p>For larger swarms, hobbling RC4 may simplify breaking the encryption
360since the same pseudorandom string is used repeatedly across the peer
361list.  Some study is in order taking into account that the tracker can
362periodically change intiailization vectors.</p>
363<p>We know from experience that periodically reshuffling peer lists on
364the order of the rerequest interval negligibly impacts tracker
365performance even with swarms containing millions of peers.  Generating
366a new pseudorandom string using RC4 on this same time interval is
367likely to incur negligible performance penalty because 1) RC4 is a
368small constant factor more expensive than a shuffle on an input string
369of equal length, 2) the generated pseudorandom string is only <em>n</em>
370ip-port pairs long where recommended <em>n</em> is within a small constant
371factor larger than the largest <em>returned peer list</em> and thus much
372smaller than the <em>tracker peer list</em> for large swarms, and 3) the cost
373of the XOR operation is lighter weight than performing a random
374shuffle.</p>
375</div>
376<div class="section" id="references">
377<h1>References</h1>
378<table class="docutils footnote" frame="void" id="mse" rules="none">
379<colgroup><col class="label" /><col /></colgroup>
380<tbody valign="top">
381<tr><td class="label">[1]</td><td><em>(<a class="fn-backref" href="#id1">1</a>, <a class="fn-backref" href="#id2">2</a>)</em> BitTorrent Message Stream Encryption
382(<a class="reference external" href="http://www.azureuswiki.com/index.php/Message_Stream_Encryption">http://www.azureuswiki.com/index.php/Message_Stream_Encryption</a>)</td></tr>
383</tbody>
384</table>
385<table class="docutils footnote" frame="void" id="borisov" rules="none">
386<colgroup><col class="label" /><col /></colgroup>
387<tbody valign="top">
388<tr><td class="label"><a class="fn-backref" href="#id3">[2]</a></td><td>Nikita Borisov, Ian Goldberg, and David Wagner. Intercepting
389mobile communications: the insecurity of 802.11. In ACM MobiCom 2001,
390pages 180-189. ACM Press, 2001.</td></tr>
391</tbody>
392</table>
393<table class="docutils footnote" frame="void" id="scott" rules="none">
394<colgroup><col class="label" /><col /></colgroup>
395<tbody valign="top">
396<tr><td class="label"><a class="fn-backref" href="#id4">[3]</a></td><td>Scott R. Fluhrer, Itsik Mantin, and Adi
397Shamir. Weaknesses in the key scheduling algorithm of RC4. In Serge
398Vaudenay and Amr M. Youssef, editors, Selected Areas in
399Cryptography 2001, volume 2259 of Lecture Notes in Computer
400Science, pages 1-24. Springer, 2001.</td></tr>
401</tbody>
402</table>
403<table class="docutils footnote" frame="void" id="stubblefeld" rules="none">
404<colgroup><col class="label" /><col /></colgroup>
405<tbody valign="top">
406<tr><td class="label"><a class="fn-backref" href="#id5">[4]</a></td><td>Adam Stubblefeld, John Ioannidis, and Aviel
407D. Rubin. A key recovery attack on the 802.11b wired equivalent
408privacy protocol (WEP). ACM Transactions on Information and System
409Security, 7(2):319-332, May 2004.</td></tr>
410</tbody>
411</table>
412<table class="docutils footnote" frame="void" id="id6" rules="none">
413<colgroup><col class="label" /><col /></colgroup>
414<tbody valign="top">
415<tr><td class="label"><a class="fn-backref" href="#id7">[5]</a></td><td><a class="reference external" href="http://tools.ietf.org/html/rfc2119">http://tools.ietf.org/html/rfc2119</a></td></tr>
416</tbody>
417</table>
418</div>
419<div class="section" id="example-python-code">
420<h1>Example Python Code</h1>
421<p>Request handling in a dummy tracker implementing tracker peer obfuscation:</p>
422<pre class="literal-block">
423from sha import sha
424from random import randint
425from struct import unpack
426from rc4 import rc4  # rc4(k) generates k RC4 pseudorandom bytes.
427
428rand = open(&quot;/dev/random&quot;,&quot;r&quot;).read
429rc4 = rc4()
430
431# tracker configuration
432MAX_PEERS = 100
433