third_party/libotr/Protocol-v3.html
author Jorg K <jorgk@jorgk.com>
Sat, 05 Oct 2019 16:18:07 +0200
changeset 35909 9ba6a8a5bcb8611538067db06d4544b1c5bdad36
parent 35648 8c34fc693509a149c5e8d490725c3b82a69acef1
permissions -rw-r--r--
No bug - Pin mozilla-esr68 (FIREFOX_68_1_0esr_RELEASE, THUNDERBIRD_68_VERBRANCH) for release. a=jorgk

<!DOCTYPE html
PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN"
"http://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd">
<html><head>
<title>Off-the-Record Messaging Protocol version 3</title>
<style type="text/css">
    body { background: white; color: black }
    h1 { text-align: center }
    dd ul.note { list-style: none }
    dl.doublespace dd { margin-bottom: 2ex }
</style>
</head><body>
<h1>Off-the-Record Messaging Protocol version 3</h1>
<p>This document describes version 3 of the Off-the-Record Messaging
protocol.  The main changes over version 2 include:</p>
<ul>
<li>Both fragmented and unfragmented messages contain sender and
recipient instance tags. This avoids an issue on IM networks that
always relay all messages to all sessions of a client who is logged
in multiple times. In this situation, OTR clients can attempt to
establish an OTR session indefinitely if there are interleaving
messages from each of the sessions.</li>
<li>An extra symmetric key is derived during AKE. This may be used for
secure communication over a different channel (e.g., file transfer,
voice chat).</li>
</ul>
<h2>Very high level overview</h2>
<p>OTR assumes a network model which provides in-order delivery of
messages, but that some messages may not get delivered at all
(for example, if the user disconnects).  There may be
an active attacker, who is allowed to perform a Denial of
Service attack, but not to learn the contents of messages.</p>
<ol>
<li>Alice signals to Bob that she would like (using an OTR Query Message)
or is willing (using a whitespace-tagged plaintext message) to use OTR
to communicate.  Either mechanism should convey the version(s) of OTR
that Alice is willing to use.</li>
<li>Bob initiates the authenticated key exchange (AKE) with Alice.
Versions 2 and 3 of OTR use a variant of the SIGMA protocol as its AKE.</li>
<li>Alice and Bob exchange Data Messages to send information to each
other.</li>
</ol>
<h2>High level overview</h2>
<h3>Requesting an OTR conversation</h3>
<p>There are two ways Alice can inform Bob that she is willing to use
the OTR protocol to speak with him: by sending him the OTR Query Message, 
or by including a special "tag" consisting of whitespace characters in
one of her messages to him.  Each method also includes a way for Alice
to communicate to Bob which versions of the OTR protocol she is willing
to speak with him.</p>
<p>The semantics of the OTR Query Message are that Alice is
<em>requesting</em> that Bob start an OTR conversation with her (if, of
course, he is willing and able to do so).  On the other hand, the
semantics of the whitespace tag are that Alice is merely
<em>indicating</em> to Bob that she is willing and able to have an OTR
conversation with him.  If Bob has a policy of "only use OTR when it's
explicitly requested", for example, then he <em>would</em> start an OTR
conversation upon receiving an OTR Query Message, but <em>would not</em>
upon receiving the whitespace tag.</p>
<h3>Authenticated Key Exchange (AKE)</h3>
<p>This section outlines the version of the SIGMA protocol used as the
AKE.  All exponentiations are done modulo a particular 1536-bit prime,
and g is a generator of that group, as indicated in the detailed
description below.  Alice and Bob's long-term authentication public keys
are pub<sub>A</sub> and pub<sub>B</sub>, respectively.</p>
<p>The general idea is that Alice and Bob do an <em>unauthenticated</em>
Diffie-Hellman (D-H) key exchange to set up an encrypted channel, and
then do mutual authentication <em>inside</em> that channel.</p>
<p>Bob will be initiating the AKE with Alice.</p>
<ul>
<li>Bob:
<ol>
<li>Picks a random value r (128 bits)</li>
<li>Picks a random value x (at least 320 bits)</li>
<li>Sends Alice AES<sub>r</sub>(g<sup>x</sup>), HASH(g<sup>x</sup>)</li>
</ol></li>
<li>Alice:
<ol>
<li>Picks a random value y (at least 320 bits)</li>
<li>Sends Bob g<sup>y</sup></li>
</ol></li>
<li>Bob:
<ol>
<li>Verifies that Alice's g<sup>y</sup> is a legal value (2 &lt;=
g<sup>y</sup> &lt;= modulus-2)</li>
<li>Computes s = (g<sup>y</sup>)<sup>x</sup></li>
<li>Computes two AES keys c, c' and four MAC keys m1, m1', m2, m2' by
hashing s in various ways</li>
<li>Picks keyid<sub>B</sub>, a serial number for his D-H key
g<sup>x</sup></li>
<li>Computes M<sub>B</sub> = MAC<sub>m1</sub>(g<sup>x</sup>, g<sup>y</sup>,
pub<sub>B</sub>, keyid<sub>B</sub>)</li>
<li>Computes X<sub>B</sub> = pub<sub>B</sub>, keyid<sub>B</sub>,
sig<sub>B</sub>(M<sub>B</sub>)</li>
<li>Sends Alice r, AES<sub>c</sub>(X<sub>B</sub>),
MAC<sub>m2</sub>(AES<sub>c</sub>(X<sub>B</sub>))</li>
</ol></li>
<li>Alice:
<ol>
<li>Uses r to decrypt the value of g<sup>x</sup> sent earlier</li>
<li>Verifies that HASH(g<sup>x</sup>) matches the value sent earlier</li>
<li>Verifies that Bob's g<sup>x</sup> is a legal value (2 &lt;=
g<sup>x</sup> &lt;= modulus-2)</li>
<li>Computes s = (g<sup>x</sup>)<sup>y</sup> (note that this will be the
same as the value of s Bob calculated)</li>
<li>Computes two AES keys c, c' and four MAC keys m1, m1', m2, m2' by
hashing s in various ways (the same as Bob)</li>
<li>Uses m2 to verify MAC<sub>m2</sub>(AES<sub>c</sub>(X<sub>B</sub>))</li>
<li>Uses c to decrypt AES<sub>c</sub>(X<sub>B</sub>) to obtain
X<sub>B</sub> = pub<sub>B</sub>, keyid<sub>B</sub>,
sig<sub>B</sub>(M<sub>B</sub>)</li>
<li>Computes M<sub>B</sub> = MAC<sub>m1</sub>(g<sup>x</sup>,
g<sup>y</sup>, pub<sub>B</sub>, keyid<sub>B</sub>)</li>
<li>Uses pub<sub>B</sub> to verify sig<sub>B</sub>(M<sub>B</sub>)</li>

<li>Picks keyid<sub>A</sub>, a serial number for her D-H key
g<sup>y</sup></li>
<li>Computes M<sub>A</sub> = MAC<sub>m1'</sub>(g<sup>y</sup>, g<sup>x</sup>,
pub<sub>A</sub>, keyid<sub>A</sub>)</li>
<li>Computes X<sub>A</sub> = pub<sub>A</sub>, keyid<sub>A</sub>,
sig<sub>A</sub>(M<sub>A</sub>)</li>
<li>Sends Bob AES<sub>c'</sub>(X<sub>A</sub>),
MAC<sub>m2'</sub>(AES<sub>c'</sub>(X<sub>A</sub>))</li>
</ol></li>
<li>Bob:
<ol>
<li>Uses m2' to verify MAC<sub>m2'</sub>(AES<sub>c'</sub>(X<sub>A</sub>))</li>
<li>Uses c' to decrypt AES<sub>c'</sub>(X<sub>A</sub>) to obtain
X<sub>A</sub> = pub<sub>A</sub>, keyid<sub>A</sub>,
sig<sub>A</sub>(M<sub>A</sub>)</li>
<li>Computes M<sub>A</sub> = MAC<sub>m1'</sub>(g<sup>y</sup>,
g<sup>x</sup>, pub<sub>A</sub>, keyid<sub>A</sub>)</li>
<li>Uses pub<sub>A</sub> to verify sig<sub>A</sub>(M<sub>A</sub>)</li>
</ol></li>
<li>If all of the verifications succeeded, Alice and Bob now know each
other's Diffie-Hellman public keys, and share the value s.  Alice is
assured that s is known by someone with access to the private key
corresponding to pub<sub>B</sub>, and similarly for Bob.</li>
</ul>
<h3>Exchanging data</h3>
<p>This section outlines the method used to protect data being exchanged
between Alice and Bob.  As above, all exponentiations are done modulo
a particular 1536-bit prime, and g is a generator of
that group, as indicated in the detailed description below.</p>
<p>Suppose Alice has a message (msg) to send to Bob.</p>
<ul>
<li>Alice:
<ol>
<li>Picks the most recent of her own D-H encryption keys that Bob has
acknowledged receiving (by using it in a Data Message, or failing that,
in the AKE).  Let key<sub>A</sub> be that key, and let keyid<sub>A</sub>
be its serial number.</li>
<li>If the above key is Alice's most recent key, she generates a new D-H key 
(next_dh), to get the serial number keyid<sub>A</sub>+1.</li>
<li>Picks the most recent of Bob's D-H encryption keys that she has
received from him (either in a Data Message or in the AKE).  Let
key<sub>B</sub> by that key, and let keyid<sub>B</sub> be its serial
number.</li>
<li>Uses Diffie-Hellman to compute a shared secret from the two keys
key<sub>A</sub> and key<sub>B</sub>, and generates the
sending AES key, ek, and the sending MAC key, mk, as detailed
below.</li>
<li>Collects any old MAC keys that were used in previous messages, but
will never again be used (because their associated D-H keys are no
longer the most recent ones) into a list, oldmackeys.</li>
<li>Picks a value of the counter, ctr, so that the triple
(key<sub>A</sub>, key<sub>B</sub>, ctr) is never the same for more
than one Data Message Alice sends to Bob.</li>
<li>Computes T<sub>A</sub> = (keyid<sub>A</sub>, keyid<sub>B</sub>, next_dh,
ctr, AES-CTR<sub>ek,ctr</sub>(msg))</li>
<li>Sends Bob T<sub>A</sub>, MAC<sub>mk</sub>(T<sub>A</sub>),
oldmackeys</li>
</ol></li>
<li>Bob:
<ol>
<li>Uses Diffie-Hellman to compute a shared secret from the two keys
labelled by keyid<sub>A</sub> and keyid<sub>B</sub>, and generates the
receiving AES key, ek, and the receiving MAC key, mk, as detailed
below.  (These will be the same as the keys Alice generated, above.)</li>
<li>Uses mk to verify MAC<sub>mk</sub>(T<sub>A</sub>).</li>
<li>Uses ek and ctr to decrypt
AES-CTR<sub>ek,ctr</sub>(msg).</li>
</ol>
</li>
</ul>
<h3>Socialist Millionaires' Protocol (SMP)</h3>
<p>While data messages are being exchanged, either Alice or Bob may
run SMP to detect impersonation or man-in-the-middle attacks.
As above, all exponentiations are done modulo a particular 1536-bit
prime, and g<sub>1</sub> is a generator of that group.  All sent values
include zero-knowledge proofs that they were generated according to
this protocol, as indicated in the detailed description below.</p>
<p>Suppose Alice and Bob have secret information x and y respectively,
and they wish to know whether x = y.  The Socialist Millionaires' Protocol
allows them to compare x and y without revealing any other information
than the value of (x == y).  For OTR, the secrets contain
information about both parties' long-term authentication public keys,
as well as information entered by the users themselves.  If x = y,
this means that Alice and Bob entered the same secret information, and
so must be the same entities who established that secret to begin with.</p>
<p>Assuming that Alice begins the exchange:</p>
<ul>
<li>Alice:
<ol>
<li>Picks random exponents a<sub>2</sub> and a<sub>3</sub></li>
<li>Sends Bob g<sub>2a</sub> = g<sub>1</sub><sup>a<sub>2</sub></sup> and
g<sub>3a</sub> = g<sub>1</sub><sup>a<sub>3</sub></sup></li>
</ol></li>
<li>Bob:
<ol>
<li>Picks random exponents b<sub>2</sub> and b<sub>3</sub></li>
<li>Computes g<sub>2b</sub> = g<sub>1</sub><sup>b<sub>2</sub></sup> and
g<sub>3b</sub> = g<sub>1</sub><sup>b<sub>3</sub></sup></li>
<li>Computes g<sub>2</sub> = g<sub>2a</sub><sup>b<sub>2</sub></sup> and
g<sub>3</sub> = g<sub>3a</sub><sup>b<sub>3</sub></sup></li>
<li>Picks random exponent r</li>
<li>Computes P<sub>b</sub> = g<sub>3</sub><sup>r</sup> and
Q<sub>b</sub> = g<sub>1</sub><sup>r</sup> g<sub>2</sub><sup>y</sup></li>
<li>Sends Alice g<sub>2b</sub>, g<sub>3b</sub>, P<sub>b</sub> and
Q<sub>b</sub></li>
</ol></li>
<li>Alice:
<ol>
<li>Computes g<sub>2</sub> = g<sub>2b</sub><sup>a<sub>2</sub></sup> and
g<sub>3</sub> = g<sub>3b</sub><sup>a<sub>3</sub></sup></li>
<li>Picks random exponent s</li>
<li>Computes P<sub>a</sub> = g<sub>3</sub><sup>s</sup> and
Q<sub>a</sub> = g<sub>1</sub><sup>s</sup> g<sub>2</sub><sup>x</sup></li>
<li>Computes R<sub>a</sub> = (Q<sub>a</sub> / Q<sub>b</sub>)
<sup>a<sub>3</sub></sup></li>
<li>Sends Bob P<sub>a</sub>, Q<sub>a</sub> and R<sub>a</sub></li>
</ol></li>
<li>Bob:
<ol>
<li>Computes R<sub>b</sub> = (Q<sub>a</sub> / Q<sub>b</sub>)
<sup>b<sub>3</sub></sup></li>
<li>Computes R<sub>ab</sub> = R<sub>a</sub><sup>b<sub>3</sub></sup></li>
<li>Checks whether R<sub>ab</sub> == (P<sub>a</sub> / P<sub>b</sub>)</li>
<li>Sends Alice R<sub>b</sub></li>
</ol></li>
<li>Alice:
<ol>
<li>Computes R<sub>ab</sub> = R<sub>b</sub><sup>a<sub>3</sub></sup></li>
<li>Checks whether R<sub>ab</sub> == (P<sub>a</sub> / P<sub>b</sub>)</li>
</ol></li>
<li>If everything is done correctly, then R<sub>ab</sub> should hold the
value of (P<sub>a</sub> / P<sub>b</sub>) times
(g<sub>2</sub><sup>a<sub>3</sub>b<sub>3</sub></sup>)<sup>(x - y)</sup>, which means that the test at the end of
the protocol will only succeed if x == y. Further, since
g<sub>2</sub><sup>a<sub>3</sub>b<sub>3</sub></sup> is a random number
not known to any party, if x is not equal to y, no other information is
revealed.</li>
</ul>
<h2>Details of the protocol</h2>
<h3>Unencoded messages</h3>
<p>This section describes the messages in the OTR protocol that are not
base-64 encoded binary.</p>
<h4>OTR Query Messages</h4>
<p>If Alice wishes to communicate to Bob that she would like to use OTR,
she sends a message containing the string "?OTR" followed by an
indication of what versions of OTR she is willing to use with Bob.  The
version string is constructed as follows:</p>
<ul>
<li>If she is willing to use OTR version 1, the version string must
start with "?".</li>
<li>If she is willing to use OTR versions other than 1, a "v" followed
by the byte identifiers for the versions in question, followed by "?".
The byte identifier for OTR version 2 is "2", and similarly for 3.  The
order of the identifiers between the "v" and the "?" does not matter,
but none should be listed more than once.</li>
</ul>
<p>For example:</p>
<dl>
<dt>"?OTR?"</dt>
<dd>Version 1 only</dd>
<dt>"?OTRv2?"</dt>
<dd>Version 2 only</dd>
<dt>"?OTRv23?"</dt>
<dd>Versions 2 and 3</dd>
<dt>"?OTR?v2?"</dt>
<dd>Versions 1 and 2</dd>
<dt>"?OTRv24x?"</dt>
<dd>Version 2, and hypothetical future versions identified by "4" and
"x"</dd>
<dt>"?OTR?v24x?"</dt>
<dd>Versions 1, 2, and hypothetical future versions identified by "4" and
"x"</dd>
<dt>"?OTR?v?"</dt>
<dd>Also version 1 only</dd>
<dt>"?OTRv?"</dt>
<dd>A bizarre claim that Alice would like to start an OTR conversation,
but is unwilling to speak any version of the protocol</dd>
</dl>
<p>These strings may be hidden from the user (for example, in
an attribute of an HTML tag), and/or may be accompanied by an
explanitory message ("Alice has requested an Off-the-Record private
conversation.").  If Bob is willing to use OTR with Alice (with a
protocol version that Alice has offered), he should start the AKE.</p>
<h4>Tagged plaintext messages</h4>
<p>If Alice wishes to communicate to Bob that she is willing to use OTR,
she can attach a special whitespace tag to any plaintext message she
sends him.  This tag may occur anywhere in the message, and may be
hidden from the user (as in the Query Messages, above).</p>
<p>The tag consists of the following 16 bytes, followed by one or more
sets of 8 bytes indicating the version of OTR Alice is willing to
use:</p>
<ul>
<li>Always send "\x20\x09\x20\x20\x09\x09\x09\x09"
"\x20\x09\x20\x09\x20\x09\x20\x20", followed by one or more of:</li>
<li>"\x20\x09\x20\x09\x20\x20\x09\x20" to indicate a willingness to use
OTR version 1 with Bob (note: this string must come before all other
whitespace version tags, if it is present, for backwards
compatibility)</li>
<li>"\x20\x20\x09\x09\x20\x20\x09\x20" to indicate a willingness to use
OTR version 2 with Bob</li>
<li>"\x20\x20\x09\x09\x20\x20\x09\x09" to indicate a willingness to use
OTR version 3 with Bob</li>
</ul>
<p>If Bob is willing to use OTR with Alice (with a protocol version that
Alice has offered), he should start the AKE.  On the other hand, if
Alice receives a plaintext message from Bob (rather than an initiation
of the AKE), she should stop sending him the whitespace tag.</p>
<h4>OTR Error Messages</h4>
<p>Any message containing the string "?OTR Error:" is an OTR Error
Message.  The following part of the message should contain
human-readable details of the error.</p>
<h3>Encoded messages</h3>
<p>This section describes the byte-level format of the base-64 encoded
binary OTR messages.  The binary form of each of the messages is
described below.  To transmit one of these messages, construct the ASCII
string consisting of the five bytes "?OTR:", followed by the base-64
encoding of the binary form of the message, followed by the byte
".".</p>
<p>For the Diffie-Hellman group computations, the group is the one
defined in RFC 3526 with 1536-bit modulus (hex, big-endian):</p>
<blockquote><pre>
FFFFFFFF FFFFFFFF C90FDAA2 2168C234 C4C6628B 80DC1CD1
29024E08 8A67CC74 020BBEA6 3B139B22 514A0879 8E3404DD
EF9519B3 CD3A431B 302B0A6D F25F1437 4FE1356D 6D51C245
E485B576 625E7EC6 F44C42E9 A637ED6B 0BFF5CB6 F406B7ED
EE386BFB 5A899FA5 AE9F2411 7C4B1FE6 49286651 ECE45B3D
C2007CB8 A163BF05 98DA4836 1C55D39A 69163FA8 FD24CF5F
83655D23 DCA3AD96 1C62F356 208552BB 9ED52907 7096966D
670C354E 4ABC9804 F1746C08 CA237327 FFFFFFFF FFFFFFFF
</pre></blockquote>
<p>and a generator (g) of 2.  Note that this means that whenever you see a
Diffie-Hellman exponentiation in this document, it always means that the
exponentiation is done modulo the above 1536-bit number.</p>
<h4>Data types</h4>
<dl>
<dt>Bytes (BYTE):</dt>
<dd>      1 byte unsigned value</dd>
<dt>Shorts (SHORT):</dt>
<dd>      2 byte unsigned value, big-endian</dd>
<dt>Ints (INT):</dt>
<dd>      4 byte unsigned value, big-endian</dd>
<dt>Multi-precision integers (MPI):</dt>
<dd>      4 byte unsigned len, big-endian
<br />      len byte unsigned value, big-endian
<br />      (MPIs must use the minimum-length encoding; i.e. no leading 0x00
      bytes.  This is important when calculating public key
      fingerprints.)</dd>
<dt>Opaque variable-length data (DATA):</dt>
<dd>      4 byte unsigned len, big-endian
<br />      len byte data</dd>
<dt>Initial CTR-mode counter value (CTR):</dt>
<dd>      8 bytes data</dd>
<dt>Message Authentication Code (MAC):</dt>
<dd>      20 bytes MAC data</dd>
</dl>
<h4>Public keys, signatures, and fingerprints</h4>
<p>OTR users have long-lived public keys that they use for
authentication (but <em>not</em> encryption).  The current version of
the OTR protocol only supports DSA public keys, but there is a key type
marker for future extensibility.</p>
<dl>
<dt>OTR public authentication DSA key (PUBKEY):</dt>
<dd>Pubkey type (SHORT)
<ul class="note"><li>DSA public keys have type 0x0000</li></ul>
p (MPI)
<br />q (MPI)
<br />g (MPI)
<br />y (MPI)
<ul class="note"><li>(p,q,g,y) are the DSA public key parameters</li></ul>
</dd>
</dl>
<p>OTR public keys are used to generate <b>signatures</b>; different
types of keys produce signatures in different formats.  The format for a
signature made by a DSA public key is as follows:</p>
<dl>
<dt>DSA signature (SIG):</dt>
<dd>      (len is the length of the DSA public parameter q, which in
current implementations must be 20 bytes, or 160 bits)
<br />      len byte unsigned r, big-endian
<br />      len byte unsigned s, big-endian</dd>
</dl>
<p>OTR public keys have <b>fingerprints</b>, which are hex strings that
serve as identifiers for the public key.  The fingerprint is calculated
by taking the SHA-1 hash of the byte-level representation of the public
key.  However, there is an exception for backwards compatibility: if the
pubkey type is 0x0000, those two leading 0x00 bytes are omitted from the
data to be hashed.  The encoding assures that, assuming the hash
function itself has no useful collisions, and DSA keys have length less
than 524281 bits (500 times larger than most DSA keys), no two public
keys will have the same fingerprint.</p>
<h4>Instance Tags</h4>
<p>Clients include instance tags in all OTR version 3 messages. Instance
tags are 32-bit values that are intended to be persistent. If the same
client is logged into the same account from multiple locations, the
intention is that the client will have different instance tags at each
location. As shown below, OTR version 3 messages (fragmented and
unfragmented) include the source and destination instance tags. If a client
receives a message that lists a destination instance tag different from its
own, the client should discard the message.</p>
<p>The smallest valid instance tag is 0x00000100. It is appropriate to set the
destination instance tag to '0' when an actual destination instance tag is
not known at the time the message is prepared. If a client receives a
message with the sender instance tag set to less than 0x00000100, it should
discard the message. Similarly, if a client receives a message with the
recipient instance tag set to greater than 0 but less than 0x00000100, it
should discard the message.
</p>

<p>This avoids an issue on IM networks that always relay all messages to
all sessions of a client who is logged in multiple times. In this
situation, OTR clients can attempt to establish an OTR session indefinitely
if there are interleaving messages from each of the sessions.</p>
<h4>D-H Commit Message</h4>
<p>This is the first message of the AKE.  Bob sends it to Alice to
commit to a choice of D-H encryption key (but the key itself is not yet
revealed).  This allows the secure session id to be much shorter than in
OTR version 1, while still preventing a man-in-the-middle attack on
it.</p>
<dl>
<dt>Protocol version (SHORT)</dt>
<dd>The version number of this protocol is 0x0003.</dd>
<dt>Message type (BYTE)</dt>
<dd>The D-H Commit Message has type 0x02.</dd>
<dt>Sender Instance tag (INT)</dt>
<dd>The instance tag of the person sending this message.</dd>
<dt>Receiver Instance tag (INT)</dt>
<dd>The instance tag of the intended recipient.
For a commit message this will often be 0, since the other party
may not have identified their instance tag yet.</dd>
<dt>Encrypted g<sup>x</sup> (DATA)</dt>
<dd>Produce this field as follows:
<ul>
<li>Choose a random value r (128 bits)</li>
<li>Choose a random value x (at least 320 bits)</li>
<li>Serialize g<sup>x</sup> as an MPI, gxmpi.  [gxmpi will probably be
196 bytes long, starting with "\x00\x00\x00\xc0".]</li>
<li>Encrypt gxmpi using AES128-CTR, with key r and initial counter value
0.  The result will be the same length as gxmpi.</li>
<li>Encode this encrypted value as the DATA field.</li>
</ul></dd>
<dt>Hashed g<sup>x</sup> (DATA)</dt>
<dd>This is the SHA256 hash of gxmpi.</dd>
</dl>
<h4>D-H Key Message</h4>
<p>This is the second message of the AKE.  Alice sends it to Bob, and it
simply consists of Alice's D-H encryption key.</p>
<dl>
<dt>Protocol version (SHORT)</dt>
<dd>The version number of this protocol is 0x0003.</dd>
<dt>Message type (BYTE)</dt>
<dd>The D-H Key Message has type 0x0a.</dd>
<dt>Sender Instance tag (INT)</dt>
<dd>The instance tag of the person sending this message.</dd>
<dt>Receiver Instance tag (INT)</dt>
<dd>The instance tag of the intended recipient.</dd>
<dt>g<sup>y</sup> (MPI)</dt>
<dd>Choose a random value y (at least 320 bits), and calculate
g<sup>y</sup>.</dd>
</dl>
<h4>Reveal Signature Message</h4>
<p>This is the third message of the AKE.  Bob sends it to Alice,
revealing his D-H encryption key (and thus opening an encrypted
channel), and also authenticating himself (and the parameters of the
channel, preventing a man-in-the-middle attack on the channel itself) to
Alice.</p>
<dl>
<dt>Protocol version (SHORT)</dt>
<dd>The version number of this protocol is 0x0003.</dd>
<dt>Message type (BYTE)</dt>
<dd>The Reveal Signature Message has type 0x11.</dd>
<dt>Sender Instance tag (INT)</dt>
<dd>The instance tag of the person sending this message.</dd>
<dt>Receiver Instance tag (INT)</dt>
<dd>The instance tag of the intended recipient.</dd>
<dt>Revealed key (DATA)</dt>
<dd>This is the value r picked earlier.</dd>
<dt>Encrypted signature (DATA)</dt>
<dd>This field is calculated as follows:
<ul>
<li>Compute the Diffie-Hellman shared secret s.</li>
<li>Use s to compute an AES key c and two MAC keys m1 and m2, as specified below.</li>
<li>Select keyid<sub>B</sub>, a serial number for the D-H key computed
earlier.  It is an INT, and must be greater than 0.</li>
<li>Compute the 32-byte value M<sub>B</sub> to be the SHA256-HMAC of the
following data, using the key m1:<dl>
<dt>g<sup>x</sup> (MPI)</dt>
<dt>g<sup>y</sup> (MPI)</dt>
<dt>pub<sub>B</sub> (PUBKEY)</dt>
<dt>keyid<sub>B</sub> (INT)</dt>
</dl></li>
<li>Let X<sub>B</sub> be the following structure:<dl>
<dt>pub<sub>B</sub> (PUBKEY)</dt>
<dt>keyid<sub>B</sub> (INT)</dt>
<dt>sig<sub>B</sub>(M<sub>B</sub>) (SIG)</dt>
<dd>This is the signature, using the private part of the key
pub<sub>B</sub>, of the 32-byte M<sub>B</sub> (taken modulo q instead of
being truncated (as described in FIPS-186), and not hashed again).</dd>
</dl></li>
<li>Encrypt X<sub>B</sub> using AES128-CTR with key c and initial
counter value 0.</li>
<li>Encode this encrypted value as the DATA field.</li>
</ul></dd>
<dt>MAC'd signature (MAC)</dt>
<dd>This is the SHA256-HMAC-160 (that is, the first 160 bits of the
SHA256-HMAC) of the encrypted signature field (including the four-byte
length), using the key m2.</dd>
</dl>
<h4>Signature Message</h4>
<p>This is the final message of the AKE.  Alice sends it to Bob,
authenticating herself and the channel parameters to him.</p>
<dl>
<dt>Protocol version (SHORT)</dt>
<dd>The version number of this protocol is 0x0003.</dd>
<dt>Message type (BYTE)</dt>
<dd>The Signature Message has type 0x12.</dd>
<dt>Sender Instance tag (INT)</dt>
<dd>The instance tag of the person sending this message.</dd>
<dt>Receiver Instance tag (INT)</dt>
<dd>The instance tag of the intended recipient.</dd>
<dt>Encrypted signature (DATA)</dt>
<dd>This field is calculated as follows:
<ul>
<li>Compute the Diffie-Hellman shared secret s.</li>
<li>Use s to compute an AES key c' and two MAC keys m1' and m2', as specified below.</li>
<li>Select keyid<sub>A</sub>, a serial number for the D-H key computed
earlier.  It is an INT, and must be greater than 0.</li>
<li>Compute the 32-byte value M<sub>A</sub> to be the SHA256-HMAC of the
following data, using the key m1':<dl>
<dt>g<sup>y</sup> (MPI)</dt>
<dt>g<sup>x</sup> (MPI)</dt>
<dt>pub<sub>A</sub> (PUBKEY)</dt>
<dt>keyid<sub>A</sub> (INT)</dt>
</dl></li>
<li>Let X<sub>A</sub> be the following structure:<dl>
<dt>pub<sub>A</sub> (PUBKEY)</dt>
<dt>keyid<sub>A</sub> (INT)</dt>
<dt>sig<sub>A</sub>(M<sub>A</sub>) (SIG)</dt>
<dd>This is the signature, using the private part of the key
pub<sub>A</sub>, of the 32-byte M<sub>A</sub> (which does not need to be
hashed again to produce the signature).</dd>
</dl></li>
<li>Encrypt X<sub>A</sub> using AES128-CTR with key c' and initial
counter value 0.</li>
<li>Encode this encrypted value as the DATA field.</li>
</ul></dd>
<dt>MAC'd signature (MAC)</dt>
<dd>This is the SHA256-HMAC-160 (that is, the first 160 bits of the
SHA256-HMAC) of the encrypted signature field (including the four-byte
length), using the key m2'.</dd>
</dl>
<h4>Data Message</h4>
<p>This message is used to transmit a private message to the
correspondent.  It is also used to reveal old MAC keys.</p>
<p>The plaintext message (either before encryption, or after decryption)
consists of a human-readable message (encoded in UTF-8, optionally with
HTML markup), optionally followed by:</p>
<ul>
<li>a single NUL (a BYTE with value 0x00), <b>and</b></li>
<li>zero or more TLV (type/length/value) records (with no padding
between them)</li>
</ul>
<p>Each TLV record is of the form:</p>
<dl>
<dt>Type (SHORT)</dt>
<dd>The type of this record.  Records with unrecognized types should be
ignored.</dd>
<dt>Length (SHORT)</dt>
<dd>The length of the following field</dd>
<dt>Value (len BYTEs)  [where len is the value of the Length field]</dt>
<dd>Any pertinent data for the record type.</dd>
</dl>
<p>Some TLV examples:</p>
<dl>
<dt>\x00\x01\x00\x00</dt>
<dd>A TLV of type 1, containing no data</dd>
<dt>\x00\x00\x00\x05\x68\x65\x6c\x6c\x6f</dt>
<dd>A TLV of type 0, containing the value "hello"</dd>
</dl>
<p>The currently defined TLV record types are:</p>
<dl>
<dt>Type 0: Padding</dt>
<dd>The value may be an arbitrary amount of data, which should be
ignored.  This type can be used to disguise the length of the plaintext
message.</dd>
<dt>Type 1: Disconnected</dt>
<dd>If the user requests to close the private connection, you may send a
message (possibly with empty human-readable part) containing a record
with this TLV type just before you discard the session keys, and
transition to MSGSTATE_PLAINTEXT (see below).  If you receive a TLV
record of this type, you should transition to MSGSTATE_FINISHED (see
below), and inform the user that his correspondent has closed his end of
the private connection, and the user should do the same.</dd>
<dt>Type 2: SMP Message 1</dt>
<dd>The value represents an initiating message of the Socialist
Millionaires' Protocol, described below.</dd>
<dt>Type 3: SMP Message 2</dt>
<dd>The value represents the second message in an instance of SMP.</dd>
<dt>Type 4: SMP Message 3</dt>
<dd>The value represents the third message in an instance of SMP.</dd>
<dt>Type 5: SMP Message 4</dt>
<dd>The value represents the final message in an instance of SMP.</dd>
<dt>Type 6: SMP Abort Message</dt>
<dd>If the user cancels SMP prematurely or encounters an error in the
protocol and cannot continue, you may send a message (possibly with empty
human-readable part) with this TLV type to instruct the other party's
client to abort the protocol.  The associated length should be zero and
the associated value should be empty.  If you receive a TLV of this type,
you should change the SMP state to SMP_EXPECT1 (see below).</dd>
<dt>Type 7: SMP Message 1Q</dt>
<dd>Like a SMP Message 1, but whose value begins with a NUL-terminated
user-specified question.</dd>
<dt>Type 8: Extra symmetric key</dt>
<dd>If you wish to use the extra symmetric key, compute it yourself as
outlined in the section "Extra symmetric key", below.  Then send this
type 8 TLV to your buddy to indicate that you'd like to use the extra
symmetric key for something.  The value of the TLV begins with a 4-byte
indication of what this symmetric key will be used for (file transfer,
voice encryption, etc.). After that, the contents are use-specific
(which file, etc.). There are no currently defined uses.  Note that the
value of the key itself is <em>not</em> placed into the TLV; your buddy
will compute it on his/her own.
</dd>
</dl>
<p>SMP Message TLVs (types 2-5) all carry data sharing the same general
format:</p> 
<dl>
<dt>MPI count (INT)</dt>
<dd>The number of MPIs contained in the remainder of the TLV.</dd>
<dt>MPI 1 (MPI)</dt>
<dd>The first MPI of the TLV, serialized into a byte array.</dd>
<dt>MPI 2 (MPI)</dt>
<dd>The second MPI of the TLV, serialized into a byte array.</dd>
<dt>etc.</dt>
</dl>
<p>There should be as many MPIs as declared in the MPI count field.  For
the exact MPIs passed for each SMP TLV, see the SMP state machine
below.</p>
<p>A message with an empty human-readable part (the plaintext is of zero
length, or starts with a NUL) is a "heartbeat" packet, and should not
be displayed to the user.  (But it's still useful to effect key
rotations.)</p>
<p>Data Message format:</p>
<dl>
<dt>Protocol version (SHORT)</dt>
<dd>The version number of this protocol is 0x0003.</dd>
<dt>Message type (BYTE)</dt>
<dd>The Data Message has type 0x03.</dd>
<dt>Sender Instance tag (INT)</dt>
<dd>The instance tag of the person sending this message.</dd>
<dt>Receiver Instance tag (INT)</dt>
<dd>The instance tag of the intended recipient.</dd>
<dt>Flags (BYTE)</dt>
<dd>The bitwise-OR of the flags for this message.  Usually you should
set this to 0x00.  The only currently defined flag is:<dl>
<dt>IGNORE_UNREADABLE (0x01)</dt>
<dd>If you receive a Data Message with this flag set, and you are unable
to decrypt the message or verify the MAC (because, for example, you
don't have the right keys), just ignore the message instead of producing
some kind of error or notification to the user.</dd>
</dl></dd>
<dt>Sender keyid (INT)</dt>
<dd>Must be strictly greater than 0, and increment by 1 with each key
change</dd>
<dt>Recipient keyid (INT)</dt>
<dd>Must therefore be strictly greater than 0, as the receiver has no
key with id 0.
<br />The sender and recipient keyids are those used to encrypt and MAC
this message.</dd>
<dt>DH y (MPI)</dt>
<dd>The *next* [i.e. sender_keyid+1] public key for the sender</dd>
<dt>Top half of counter init (CTR)</dt>
<dd>This should monotonically increase (as a big-endian value) for
      each message sent with the same (sender keyid, recipient keyid)
      pair, and must not be all 0x00.</dd>
<dt>Encrypted message (DATA)</dt>
<dd>Using the appropriate encryption key (see below) derived from the
      sender's and recipient's DH public keys (with the keyids given in
      this message), perform AES128 counter-mode (CTR) encryption of the
      message.  The initial counter is a 16-byte value whose first 8
      bytes are the above "top half of counter init" value, and whose
      last 8 bytes are all 0x00.  Note that counter mode does not change
      the length of the message, so no message padding needs to be done.
      If you *want* to do message padding (to disguise the length of
      your message), use the above TLV of type 0.</dd>
<dt>Authenticator (MAC)</dt>
<dd>The SHA1-HMAC, using the appropriate MAC key (see below) of everything
    from the Protocol version to the end of the encrypted message</dd>
<dt>Old MAC keys to be revealed (DATA)</dt>
<dd>See "Revealing MAC Keys", below.</dd>
</dl>
<h3>Socialist Millionaires' Protocol (SMP)</h3>
<p>The Socialist Millionaires' Protocol allows two parties with secret
information x and y respectively to check whether (x==y) without revealing
any additional information about the secrets.  The protocol used by OTR is
based on the work of Boudot, Schoenmakers and Traore (2001).  A full 
justification for its use in OTR is made by Alexander and Goldberg,
in a paper published in 2007.  The following is a technical account
of what is transmitted during the course of the protocol.</p>
<h4>Secret information</h4>
<p>The secret information x and y compared during this protocol contains
not only information entered by the users, but also information unique to
the conversation in which SMP takes place.  Specifically, the format is:</p>
<dl>
<dt>Version (BYTE)</dt>
<dd>The version of SMP used.  The version described here is 1.</dd>
<dt>Initiator fingerprint (20 BYTEs)</dt>
<dd>The fingerprint that the party initiating SMP is using in
the current conversation.</dd>
<dt>Responder fingerprint (20 BYTEs)</dt>
<dd>The fingerprint that the party that did not initiate SMP is 
using in the current conversation.</dd>
<dt>Secure Session ID</dt>
<dd>The ssid described below.</dd>
<dt>User-specified secret</dt>
<dd>The input string given by the user at runtime.</dd>
</dl>
<p>Then the SHA256 hash of the above is taken, and the digest becomes the 
actual secret (x or y) to be used in SMP.  The additional fields insure
that not only do both parties know the same secret input string, but no
man-in-the-middle is capable of reading their communication either.</p>
<h3>The SMP state machine</h3>
<p>Whenever the OTR message state machine has MSGSTATE_ENCRYPTED set
(see below), the SMP state machine may progress.  If at any point 
MSGSTATE_ENCRYPTED becomes unset, SMP must abandon its state and return
to its initial setup.  The SMP state consists of one main variable, as
well as information from the partial computations at each protocol step.</p>
<h4>Expected Message</h4>
<p>This main state variable for SMP controls what SMP-specific TLVs will
be accepted.  This variable has no effect on type 0 or type 1 TLVs, which
are always allowed.  smpstate can take one of four values:</p>
<dl>
<dt>SMPSTATE_EXPECT1</dt>
<dd>This state indicates that only type 2 (SMP message 1) and type 7
(SMP message 1Q) TLVs should be accepted.  This is the default state when
SMP has not yet begun.  This state is also reached whenever an error
occurs or SMP is aborted, and the protocol must be restarted from the
beginning.</dd>
<dt>SMPSTATE_EXPECT2</dt>
<dd>This state indicates that only type 3 TLVs (SMP message 2) should
be accepted.</dd>
<dt>SMPSTATE_EXPECT3</dt>
<dd>This state indicates that only type 4 TLVs (SMP message 3) should
be accepted.</dd>
<dt>SMPSTATE_EXPECT4</dt>
<dd>This state indicates that only type 5 TLVs (SMP message 4) should
be accepted.</dd>
</dl>
<h4>State Transitions</h4>
<p>There are 7 actions that an OTR client must handle:</p>
<ul>
<li>Received TLVs:
<ul>
<li>SMP Message 1</li>
<li>SMP Message 2</li>
<li>SMP Message 3</li>
<li>SMP Message 4</li>
<li>SMP Abort Message</li>
</ul></li>
<li>User actions:</li>
<ul>
<li>User requests to begin SMP</li>
<li>User requests to abort SMP</li>
</ul></li>
</ul>
<p>The following sections outline what is to be done in each case.  They
all assume that MSGSTATE_ENCRYPTED is set.  For simplicity, they also
assume that Alice has begun SMP, and Bob is responding to her.</p>
<h4>SMP Hash function</h4>
<p>In the following actions, there are many places where a SHA256 hash of
an integer followed by one or two MPIs is taken.  The input to this hash
function is:</p>
<dl>
<dt>Version (BYTE)</dt>
<dd>This distinguishes calls to the hash function at different points in
the protocol, to prevent Alice from replaying Bob's zero knowledge proofs
or vice versa.</dd>
<dt>First MPI (MPI)</dt>
<dd>The first MPI given as input, serialized in the usual way.</dd>
<dt>Second MPI (MPI)</dt>
<dd>The second MPI given as input, if present, serialized in the usual way.
If only one MPI is given as input, this field is simply omitted.</dd>
</dl>
<h4>Receiving a type 2 TLV (SMP message 1)</h4>
<p>SMP message 1 is sent by Alice to begin a DH exchange to determine two
new generators, g<sub>2</sub> and g<sub>3</sub>.  It contains the
following mpi values:</p>
<dl>
<dt>g<sub>2a</sub></dt>
<dd>Alice's half of the DH exchange to determine g<sub>2</sub>.</dd>
<dt>c2, D2</dt>
<dd>A zero-knowledge proof that Alice knows the exponent associated with
her transmitted value g<sub>2a</sub>.</dd>
<dt>g<sub>3a</sub></dt>
<dd>Alice's half of the DH exchange to determine g<sub>3</sub>.</dd>
<dt>c3, D3</dt>
<dd>A zero-knowledge proof that Alice knows the exponent associated with
her transmitted value g<sub>3a</sub>.</dd>
</dl>
<p>A type 7 (SMP Message 1Q) TLV is the same as the above, but is
preceded by a user-specified question, which is associated with the
user-specified portion of the secret.</p>
<p>When Bob receives this TLV he should do:</p>
<dl>
<dt>If smpstate is not SMPSTATE_EXPECT1:</dt>
<dd>Set smpstate to SMPSTATE_EXPECT1 and send a type 6 TLV (SMP abort)
to Alice.</dd>
<dt>If smpstate is SMPSTATE_EXPECT1:</dt>
<dd>Verify Alice's zero-knowledge proofs for g<sub>2a</sub> and
g<sub>3a</sub>:
<ol>
<li>Check that both g<sub>2a</sub> and g<sub>3a</sub> are &gt;= 2 and
&lt;= modulus-2.</li>
<li>Check that c2 = SHA256(1, g<sub>1</sub><sup>D2</sup> 
g<sub>2a</sub><sup>c2</sup>).</li>
<li>Check that c3 = SHA256(2, g<sub>1</sub><sup>D3</sup> 
g<sub>3a</sub><sup>c3</sup>).</li>
</ol>
Create a type 3 TLV (SMP message 2) and send it to Alice:
<ol>
<li>Determine Bob's secret input y, which is to be compared to Alice's
secret x.</li>
<li>Pick random exponents b<sub>2</sub> and b<sub>3</sub>.
These will used during the DH exchange to pick generators.</li>
<li>Pick random exponents r2, r3, r4, r5 and r6.
These will be used to add a blinding factor to the final results, and
to generate zero-knowledge proofs that this message was created honestly.</li>
<li>Compute g<sub>2b</sub> = g<sub>1</sub><sup>b<sub>2</sub></sup> and
g<sub>3b</sub> = g<sub>1</sub><sup>b<sub>3</sub></sup></li>
<li>Generate a zero-knowledge proof that the exponent b<sub>2</sub> is
known by setting c2 = SHA256(3, g<sub>1</sub><sup>r2</sup>) and
D2 = r2 - b<sub>2</sub> c2 mod q. In the zero-knowledge proofs the D values
are calculated modulo q = (p - 1) / 2, where p is the same 1536-bit prime
as elsewhere. The random exponents are 1536-bit numbers.</li>
<li>Generate a zero-knowledge proof that the exponent b<sub>3</sub> is
known by setting c3 = SHA256(4, g<sub>1</sub><sup>r3</sup>) and
D3 = r3 - b<sub>3</sub> c3 mod q.</li>
<li>Compute g<sub>2</sub> = g<sub>2a</sub><sup>b<sub>2</sub></sup> and
g<sub>3</sub> = g<sub>3a</sub><sup>b<sub>3</sub></sup></li>
<li>Compute P<sub>b</sub> = g<sub>3</sub><sup>r4</sup> and
Q<sub>b</sub> = g<sub>1</sub><sup>r4</sup> g<sub>2</sub><sup>y</sup></li>
<li>Generate a zero-knowledge proof that P<sub>b</sub> and Q<sub>b</sub>
were created according to the protocol by setting 
cP = SHA256(5, g<sub>3</sub><sup>r5</sup>, g<sub>1</sub><sup>r5</sup> 
g<sub>2</sub><sup>r6</sup>), D5 = r5 - r4 cP mod q and D6 = r6 - y cP mod q.</li>
<li>Store the values of g<sub>3a</sub>, g<sub>2</sub>, g<sub>3</sub>,
b<sub>3</sub>, P<sub>b</sub> and Q<sub>b</sub> for use later in the
protocol.</li>
<li>Send Alice a type 3 TLV (SMP message 2) containing g<sub>2b</sub>, 
c2, D2, g<sub>3b</sub>, c3, D3, P<sub>b</sub>, Q<sub>b</sub>, cP, D5
and D6, in that order.</li>
</ol>
Set smpstate to SMPSTATE_EXPECT3.</dd>
</dl>
<h4>Receiving a type 3 TLV (SMP message 2)</h4>
<p>SMP message 2 is sent by Bob to complete the DH exchange to
determine the new generators, g<sub>2</sub> and g<sub>3</sub>. 
It also begins the construction of the values used in the final
comparison of the protocol.  It contains the following mpi values:</p>
<dl>
<dt>g<sub>2b</sub></dt>
<dd>Bob's half of the DH exchange to determine g<sub>2</sub>.</dd>
<dt>c2, D2</dt>
<dd>A zero-knowledge proof that Bob knows the exponent associated with
his transmitted value g<sub>2b</sub>.</dd>
<dt>g<sub>3b</sub></dt>
<dd>Bob's half of the DH exchange to determine g<sub>3</sub>.</dd>
<dt>c3, D3</dt>
<dd>A zero-knowledge proof that Bob knows the exponent associated with
his transmitted value g<sub>3b</sub>.</dd>
<dt>P<sub>b</sub>, Q<sub>b</sub></dt>
<dd>These values are used in the final comparison to determine if Alice
and Bob share the same secret.</dd>
<dt>cP, D5, D6</dt>
<dd>A zero-knowledge proof that P<sub>b</sub> and Q<sub>b</sub> were
created according to the protcol given above.</dd>
</dl>
<p>When Alice receives this TLV she should do:</p>
<dl>
<dt>If smpstate is not SMPSTATE_EXPECT2:</dt>
<dd>Set smpstate to SMPSTATE_EXPECT1 and send a type 6 TLV (SMP abort)
to Bob.</dd>
<dt>If smpstate is SMPSTATE_EXPECT2:</dt>
<dd>Verify Bob's zero-knowledge proofs for g<sub>2b</sub>, 
g<sub>3b</sub>, P<sub>b</sub> and Q<sub>b</sub>:
<ol>
<li>Check that g<sub>2b</sub>,
g<sub>3b</sub>, P<sub>b</sub> and Q<sub>b</sub> are &gt;= 2 and
&lt;= modulus-2.</li>
<li>Check that c2 = SHA256(3, g<sub>1</sub><sup>D2</sup> 
g<sub>2b</sub><sup>c2</sup>).</li>
<li>Check that c3 = SHA256(4, g<sub>1</sub><sup>D3</sup> 
g<sub>3b</sub><sup>c3</sup>).</li>
<li>Check that cP = SHA256(5, g<sub>3</sub><sup>D5</sup> 
P<sub>b</sub><sup>cP</sup>, g<sub>1</sub><sup>D5</sup>
g<sub>2</sub><sup>D6</sup> Q<sub>b</sub><sup>cP</sup>).</li>
</ol>
Create a type 4 TLV (SMP message 3) and send it to Bob:
<ol>
<li>Pick random exponents r4, r5, r6 and r7.
These will be used to add a blinding factor to the final results, and
to generate zero-knowledge proofs that this message was created honestly.</li>
<li>Compute g<sub>2</sub> = g<sub>2b</sub><sup>a<sub>2</sub></sup> and
g<sub>3</sub> = g<sub>3b</sub><sup>a<sub>3</sub></sup></li>
<li>Compute P<sub>a</sub> = g<sub>3</sub><sup>r4</sup> and
Q<sub>a</sub> = g<sub>1</sub><sup>r4</sup> g<sub>2</sub><sup>x</sup></li>
<li>Generate a zero-knowledge proof that P<sub>a</sub> and Q<sub>a</sub>
were created according to the protocol by setting 
cP = SHA256(6, g<sub>3</sub><sup>r5</sup>, g<sub>1</sub><sup>r5</sup> 
g<sub>2</sub><sup>r6</sup>), D5 = r5 - r4 cP mod q and D6 = r6 - x cP mod q.</li>
<li>Compute R<sub>a</sub> = (Q<sub>a</sub> / Q<sub>b</sub>)
<sup>a<sub>3</sub></sup></li>
<li>Generate a zero-knowledge proof that R<sub>a</sub> was created 
according to the protocol by setting cR = SHA256(7, g<sub>1</sub><sup>r7</sup>, 
(Q<sub>a</sub> / Q<sub>b</sub>)<sup>r7</sup>) and 
D7 = r7 - a<sub>3</sub> cR mod q.</li>
<li>Store the values of g<sub>3b</sub>, (P<sub>a</sub> / P<sub>b</sub>), 
(Q<sub>a</sub> / Q<sub>b</sub>) and a<sub>3</sub> for use later in the
protocol.</li>
<li>Send Bob a type 4 TLV (SMP message 3) containing P<sub>a</sub>, 
Q<sub>a</sub>, cP, D5, D6, R<sub>a</sub>, cR and D7 in that order.</li>
</ol>
Set smpstate to SMPSTATE_EXPECT4.</dd>
</dl>
<h4>Receiving a type 4 TLV (SMP message 3)</h4>
<p>SMP message 3 is Alice's final message in the SMP exchange.  It 
has the last of the information required by Bob to determine if x = y.
It contains the following mpi values:</p>
<dl>
<dt>P<sub>a</sub>, Q<sub>a</sub></dt>
<dd>These values are used in the final comparison to determine if Alice
and Bob share the same secret.</dd>
<dt>cP, D5, D6</dt>
<dd>A zero-knowledge proof that P<sub>a</sub> and Q<sub>a</sub> were
created according to the protcol given above.</dd>
<dt>R<sub>a</sub></dt>
<dd>This value is used in the final comparison to determine if Alice
and Bob share the same secret.</dd>
<dt>cR, D7</dt>
<dd>A zero-knowledge proof that R<sub>a</sub> was
created according to the protcol given above.</dd>
<dt>
</dl>
<p>When Bob receives this TLV he should do:</p>
<dl>
<dt>If smpstate is not SMPSTATE_EXPECT3:</dt>
<dd>Set smpstate to SMPSTATE_EXPECT1 and send a type 6 TLV (SMP abort)
to Bob.</dd>
<dt>If smpstate is SMPSTATE_EXPECT3:</dt>
<dd>Verify Alice's zero-knowledge proofs for P<sub>a</sub>, Q<sub>a</sub> 
and R<sub>a</sub>:
<ol>
<li>Check that P<sub>a</sub>, Q<sub>a</sub> and R<sub>a</sub> are &gt;= 2 and
&lt;= modulus-2.</li>
<li>Check that cP = SHA256(6, g<sub>3</sub><sup>D5</sup> 
P<sub>a</sub><sup>cP</sup>, g<sub>1</sub><sup>D5</sup> g<sub>2</sub><sup>D6</sup> 
Q<sub>a</sub><sup>cP</sup>).</li>
<li>Check that cR = SHA256(7, g<sub>1</sub><sup>D7</sup>
g<sub>3a</sub><sup>cR</sup>, (Q<sub>a</sub> / Q<sub>b</sub>)<sup>D7</sup> 
R<sub>a</sub><sup>cR</sup>).</li> 
</ol>
Create a type 5 TLV (SMP message 4) and send it to Alice:
<ol>
<li>Pick a random exponent r7.
This will be used to generate Bob's final zero-knowledge proof that 
this message was created honestly.</li>
<li>Compute R<sub>b</sub> = (Q<sub>a</sub> / Q<sub>b</sub>)
<sup>b<sub>3</sub></sup></li>
<li>Generate a zero-knowledge proof that R<sub>b</sub> was created 
according to the protocol by setting cR = SHA256(8, g<sub>1</sub><sup>r7</sup>, 
(Q<sub>a</sub> / Q<sub>b</sub>)<sup>r7</sup>) and 
D7 = r7 - b<sub>3</sub> cR mod q.</li>
<li>Send Alice a type 5 TLV (SMP message 4) containing R<sub>b</sub>, 
cR and D7 in that order.</li>
</ol>
Check whether the protocol was successful:
<ol>
<li>Compute R<sub>ab</sub> = R<sub>a</sub><sup>b<sub>3</sub></sup>.</li>
<li>Determine if x = y by checking the equivalent condition that 
(P<sub>a</sub> / P<sub>b</sub>) = R<sub>ab</sub>.</li>
</ol>
Set smpstate to SMPSTATE_EXPECT1, as no more messages are expected from
Alice.</dd>
</dl>
<h4>Receiving a type 5 TLV (SMP message 4)</h4>
<p>SMP message 4 is Bob's final message in the SMP exchange.  It 
has the last of the information required by Alice to determine if x = y.
It contains the following mpi values:</p>
<dl>
<dt>R<sub>b</sub></dt>
<dd>This value is used in the final comparison to determine if Alice
and Bob share the same secret.</dd>
<dt>cR, D7</dt>
<dd>A zero-knowledge proof that R<sub>b</sub> was
created according to the protcol given above.</dd>
<dt>
</dl>
<p>When Alice receives this TLV she should do:</p>
<dl>
<dt>If smpstate is not SMPSTATE_EXPECT4:</dt>
<dd>Set smpstate to SMPSTATE_EXPECT1 and send a type 6 TLV (SMP abort)
to Bob.</dd>
<dt>If smpstate is SMPSTATE_EXPECT4:</dt>
<dd>Verify Bob's zero-knowledge proof for R<sub>b</sub>:
<ol>
<li>Check that R<sub>b</sub> is &gt;= 2 and
&lt;= modulus-2.</li>
<li>Check that cR = SHA256(8, g<sub>1</sub><sup>D7</sup>
g<sub>3b</sub><sup>cR</sup>, (Q<sub>a</sub> / Q<sub>b</sub>)<sup>D7</sup> 
R<sub>b</sub><sup>cR</sup>).</li> 
</ol>
Check whether the protocol was successful:
<ol>
<li>Compute R<sub>ab</sub> = R<sub>b</sub><sup>a<sub>3</sub></sup>.</li>
<li>Determine if x = y by checking the equivalent condition that 
(P<sub>a</sub> / P<sub>b</sub>) = R<sub>ab</sub>.</li>
</ol>
Set smpstate to SMPSTATE_EXPECT1, as no more messages are expected from
Bob.</dd>
</dl>
<h4>User requests to begin SMP</h4>
<dl>
<dt>If smpstate is not set to SMPSTATE_EXPECT1:</dt>
<dd>SMP is already underway.  If you wish to restart SMP, send a
type 6 TLV (SMP abort) to the other party and then proceed as if
smpstate was SMPSTATE_EXPECT1.  Otherwise, you may simply continue the
current SMP instance.</dd>
<dt>If smpstate is set to SMPSTATE_EXPECT1:</dt>
<dd>No current exchange is underway.  In this case, Alice should 
create a valid type 2 TLV (SMP message 1) as follows:
<ol>
<li>Determine her secret input x, which is to be compared to Bob's
secret y.</li>
<li>Pick random values a<sub>2</sub> and a<sub>3</sub> (1536 bits).
These will be Alice's exponents for the DH exchange to pick generators.</li>
<li>Pick random values r2 and r3 (1536 bits).
These will be used to generate zero-knowledge proofs that this message
was created according to the protocol.</li>
<li>Compute g<sub>2a</sub> = g<sub>1</sub><sup>a<sub>2</sub></sup> and
g<sub>3a</sub> = g<sub>1</sub><sup>a<sub>3</sub></sup></li>
<li>Generate a zero-knowledge proof that the exponent a<sub>2</sub> is
known by setting c2 = SHA256(1, g<sub>1</sub><sup>r2</sup>) and
D2 = r2 - a<sub>2</sub> c2 mod q.</li>
<li>Generate a zero-knowledge proof that the exponent a<sub>3</sub> is
known by setting c3 = SHA256(2, g<sub>1</sub><sup>r3</sup>) and
D3 = r3 - a<sub>3</sub> c3 mod q.</li>
<li>Store the values of x, a<sub>2</sub> and a<sub>3</sub> 
for use later in the protocol.</li>
<li>Send Bob a type 2 TLV (SMP message 1) containing g<sub>2a</sub>, 
c2, D2, g<sub>3a</sub>, c3 and D3 in that order.</li>
</ol>
Set smpstate to SMPSTATE_EXPECT2.</dd>
</dl>
<h4>User requests to abort SMP</h4>
<p>In all cases, send a type 6 TLV (SMP abort) to the correspondent and
set smpstate to SMPSTATE_EXPECT1.</p>
<h3>Key Management</h3>
<p>For each correspondent, keep track of:</p>
<dl>
<dt>Your two most recent DH public/private key pairs</dt>
<dd>our_dh[our_keyid] (most recent) and our_dh[our_keyid-1] (previous)</dd>
<dt>His two most recent DH public keys</dt>
<dd>their_y[their_keyid] (most recent) and their_y[their_keyid-1]
(previous)</dd>
</dl>

<p>When starting a private conversation with a correspondent, generate
two DH key pairs for yourself, and set our_keyid = 2.  Note that all DH
key pairs should have a private part that is at least 320 bits long.</p>

<dl class="doublespace">
<dt>When you send AKE messages:</dt>
<dd>Send the public part of our_dh[our_keyid-1], with the keyid field,
    of course, set to (our_keyid-1).</dd>

<dt>Upon completing the AKE:</dt>
<dd>If the specified keyid equals either their_keyid or their_keyid-1,
    and the DH pubkey contained in the AKE messages matches the
    one you've stored for that keyid, that's great.  Otherwise, forget
    all values of their_y[], and of their_keyid, and set their_keyid to
    the keyid value given in the AKE messages, and
    their_y[their_keyid] to the DH pubkey value given in the AKE
    messages.  their_y[their_keyid-1] should be set to NULL.</dd>

<dt>When you send a Data Message:</dt>
<dd>Set the sender keyid to (our_keyid-1), and the recipient keyid to
    (their_keyid).  Set the DH pubkey in the Data message to the public
    part of our_dh[our_keyid].  Use our_dh[our_keyid-1] and
    their_y[their_keyid] to calculate session keys, as outlined below.
    Use the "sending AES key" to encrypt the message, and the "sending
    MAC key" to calculate its MAC.</dd>

<dt>When you receive a Data Message:</dt>
<dd>Use the keyids in the message to select which of your DH key pairs
    and which of his DH pubkeys to use to verify the MAC.  If the keyids
    do not represent either the most recent key or the previous key (for
    either the sender or receiver), reject the message.  Also reject the
    message if the sender keyid is their_keyid-1, but
    their_y[their_keyid-1] is NULL.

    <p>Otherwise, calculate the session keys as outlined below.  Use the
    "receiving MAC key" to verify the MAC on the message.  If it does not
    verify, reject the message.</p>

    <p>Check that the counter in the Data message is strictly larger than the
    last counter you saw using this pair of keys.  If not, reject the
    message.</p>

    <p>If the MAC verifies, decrypt the message using the "receiving AES
    key".</p>

    <p>Finally, check if keys need rotation:</p>
    <ul>
    <li>If the "recipient keyid" in the Data message equals our_keyid, then
       he's seen the public part of our most recent DH key pair, so you
       must securely forget our_dh[our_keyid-1], increment our_keyid, and set
       our_dh[our_keyid] to a new DH key pair which you generate.</li>
    <li>If the "sender keyid" in the Data message equals their_keyid,
       increment their_keyid, and set their_y[their_keyid] to the new DH
       pubkey specified in the Data message.</li>
    </ul></dd>
</dl>

<h4>Computing AES keys, MAC keys, and the secure session id</h4>
<p>OTR uses Diffie-Hellman to calculate shared secrets in the usual way:
if Bob knows x, and tells Alice g<sup>x</sup>, and Alice knows y, and
tells Bob g<sup>y</sup>, then they each can calculate s =
g<sup>xy</sup>: Alice calculates (g<sup>x</sup>)<sup>y</sup>, and Bob
calculates (g<sup>y</sup>)<sup>x</sup>.</p>
<p>During the AKE, Alice and Bob each calculate s in this way, and then
they each compute seven values based on s:</p>
<ul>
<li>A 64-bit secure session id, ssid</li>
<li>Two 128-bit AES encryption keys, c and c'</li>
<li>Four 256-bit SHA256-HMAC keys, m1, m2, m1', and m2'</li>
</ul>
<p>This is done in the following way:</p>
<ul>
<li>Write the value of s as a minimum-length MPI, as specified above
(4-byte big-endian len, len-byte big-endian value).  Let this
(4+len)-byte value be "secbytes".</li>
<li>For a given byte b, define h2(b) to be the 256-bit output of the
SHA256 hash of the (5+len) bytes consisting of the byte b followed by
secbytes.</li>
<li>Let ssid be the first 64 bits of h2(0x00).</li>
<li>Let c be the first 128 bits of h2(0x01), and let c' be the second
128 bits of h2(0x01).</li>
<li>Let m1 be h2(0x02).</li>
<li>Let m2 be h2(0x03).</li>
<li>Let m1' be h2(0x04).</li>
<li>Let m2' be h2(0x05).</li>
</ul>
<p>c, m1, and m2 are used to create and verify the Reveal Signature
Message; c', m1', and m2' are used to create and verify the Signature
message.</p>
<p>If the user requests to see the secure session id, it should be
displayed as two 32-bit bigendian unsigned values, in C "%08x" format.
If the user transmitted the Reveal Signature message during the AKE that
produced this ssid, then display the first 32 bits in bold, and the
second 32 bits in non-bold.  If the user transmitted the Signature
message instead, display the first 32 bits in non-bold, and the
second 32 bits in bold.  This session id can be used by the parties to
verify (say, over the telephone, assuming the parties recognize each
others' voices) that there is no man-in-the-middle by having each side
read his bold part to the other.  [Note that this only needs to be done
in the event that the users do not trust that their long-term signature
keys have not been compromised.]</p>
<p>During the exchange of Data Messages, Alice and Bob use the keyids
listed in the Data Message to select Diffie-Hellman keys to use to
compute s, and the (4+len)-byte value of secbytes, as above.</p>
<p>From this, they calculate four values:</p>
<ul>
<li>Two 128-bit AES encryption keys, the "sending AES key", and the
"receiving AES key"</li>
<li>Two 160-bit SHA1-HMAC keys, the "sending MAC key", and the
"receiving MAC key"</li>
</ul>
<p>These keys are calculated as follows:</p>
<ul>
<li>Alice (and similarly for Bob) determines if she is the "low" end
or the "high" end of this Data Message.  If Alice's public key is
numerically greater than Bob's public key, then she
is the "high" end.  Otherwise, she is the "low" end.  Note that who is the
"low" end and who is the "high" end can change every time a new D-H
public key is exchanged in a Data Message.</li>
<li>She sets the values of "sendbyte" and "recvbyte" according to
whether she is the the "low" or the "high" end of the Data Message:
<ul>
<li>If she is the "high" end, she sets "sendbyte" to 0x01 and "recvbyte"
to 0x02.</li>
<li>If she is the "low" end, she sets "sendbyte" to 0x02 and "recvbyte"
to 0x01.</li>
</ul></li>
<li>For a given byte b, define h1(b) to be the 160-bit output of the
SHA-1 hash of the (5+len) bytes consisting of the byte b, followed by
secbytes.</li>
<li>The "sending AES key" is the first 16 bytes of h1(sendbyte).</li>
<li>The "sending MAC key" is the 20-byte SHA-1 hash of the 16-byte
sending AES key.</li>
<li>The "receiving AES key" is the first 16 bytes of h1(recvbyte).</li>
<li>The "receiving MAC key" is the 20-byte SHA-1 hash of the 16-byte
receiving AES key.</li>
</ul>
<h4>Extra symmetric key</h4>
<p>OTR version 3 defines an additional symmetric key that can be derived
by the communicating parties to use for application-specific purposes,
such as file transfer, voice encryption, etc.  When one party wishes to
use the extra symmetric key, he or she creates a type 8 TLV attached to
a Data Message (see above).  The key itself is then derived using the
same "secbytes" used to compute the encryption and MAC keys used to
protect the Data Message.
The extra symmetric key is derived by calculating
h2(0xFF) and keeping the entire 256 bits, using the same definition
of h2 as above.</p>
<p>Upon receipt of the Data Message containing the type 8 TLV, the
recipient will compute the extra symmetric key in the same way.  Note
that the value of the extra symmetric key is <em>not</em> contained in
the TLV itself.</p>
<h4>Revealing MAC keys</h4>
<p>Whenever you are about to forget either one of your old D-H key pairs, or
one of your correspondent's old D-H public keys, take all of the
receiving MAC keys
that were generated by that key (note that there are up to two: the
receiving MAC keys produced by the pairings of that key with
each of two of the other side's keys; but note that you only need to
take MAC keys that were actually used to verify a MAC on a message), and
put them (as a set of
concatenated 20-byte values) into the "Old MAC keys to be revealed"
section of the next Data Message you send.  This in done to allow the
forgeability of OTR transcripts: once the MAC keys are revealed, anyone
can modify an OTR message and still have it appear valid.  But since we
don't reveal the MAC keys until their corresponding pubkeys are being
discarded, there is no danger of accepting a message as valid which
uses a MAC key which has already been revealed.</p>
<h3>Fragmentation</h3>
<p>Some networks may have a maximum message size that is too small to
contain an encoded OTR message.  In that event, the sender may choose
to split the message into a number of <em>fragments</em>.  This section
describes the format of the fragments.  All OTR version 2 and 3 clients
must be able to assemble received fragments, but performing
fragmentation on outgoing messages is optional.</p>

<dl class="doublespace">
<dt>Transmitting Fragments</dt>
<dd>If you have information about the maximum size of message you are
    able to send (the different IM networks have different limits), you
    can fragment an encoded OTR message as follows:
    <ul>
    <li>Start with the OTR message as you would normally transmit it.  For
      example, a Data Message would start with "?OTR:AAED" and end
      with ".".</li>
    <li>Break it up into sufficiently small pieces.  Let the number of
      pieces be (n), and the pieces be
      piece[1],piece[2],...,piece[n].</li>
    <li>Transmit (n) OTR version 3 fragmented messages with the following
      (printf-like) structure (as k runs from 1 to n inclusive):

      <p>"?OTR|%x|%x,%hu,%hu,%s," , sender_instance, receiver_instance, 
      k , n , piece[k]</p>
      
      OTR version 2 messages get fragmented in a similar format, but 
      without the instance tags fields:
      
      <p>"?OTR,%hu,%hu,%s," ,
      k , n , piece[k]</p></li>

    <li>Note that k and n are unsigned short ints (2 bytes), and each has
      a maximum value of 65535.  Also, each piece[k] must be
      non-empty.  The instance tags (if applicable) and the k and n
      values may have leading zeroes.</li>
    </ul>
    <p>Note that fragments are not themselves messages that can be
    fragmented: you can't fragment a fragment.</p></dd>

<dt>Receiving Fragments:</dt>

<dd>If you receive a message containing "?OTR|" (note that you'll need
    to check for this _before_ checking for any of the other "?OTR:"
    markers):

    <ul>
    <li>Parse it as the printf statement above into k, n, and
    piece.</li>
    <li>If the recipient's own instance tag does not match the listed
    receiver instance tag, and the listed receiver instance tag is not
    zero, the recipient should discard the message and optionally pass
    along a warning for the user.</li>
    <li>Let (K,N) be your currently stored fragment number, and F be your
      currently stored fragment.  [If you have no currently stored
      fragment, then K = N = 0 and F = "".]</li>

    <li>If k == 0 or n == 0 or k &gt; n, discard this (illegal)
    fragment.</li>

    <li>If k == 1:
    <ul>
      <li>Forget any stored fragment you may have</li>
      <li>Store (piece) as F.</li>
      <li>Store (k,n) as (K,N).</li>
    </ul></li>

    <li>If n == N and k == K+1:
    <ul>
      <li>Append (piece) to F.</li>
      <li>Store (k,n) as (K,N).</li>
    </ul></li>

    <li>Otherwise:
    <ul>
      <li>Forget any stored fragment you may have</li>
      <li>Store "" as F.</li>
      <li>Store (0,0) as (K,N).</li>
    </ul></li>
    </ul>

    <p>After this, if N &gt; 0 and K == N, treat F as the received
    message.</p>

    <p>If you receive a non-OTR message, or an unfragmented message,
    forget any stored fragment you may have, store "" as F and store
    (0,0) as (K,N).</p>
    
    <p>OTR version 2 fragmented messages follow the same behaviour as
    described above, but do not list the sender and receiver instance
    tags.</dd>
</dl>

<p>For example, here is a Data Message we would like to transmit over a
network with an unreasonably small maximum message size:</p>

<blockquote><pre>
?OTR:AAMDJ+MVmSfjFZcAAAAAAQAAAAIAAADA1g5IjD1ZGLDVQEyCgCyn9hb
rL3KAbGDdzE2ZkMyTKl7XfkSxh8YJnudstiB74i4BzT0W2haClg6dMary/jo
9sMudwmUdlnKpIGEKXWdvJKT+hQ26h9nzMgEditLB8vjPEWAJ6gBXvZrY6ZQ
rx3gb4v0UaSMOMiR5sB7Eaulb2Yc6RmRnnlxgUUC2alosg4WIeFN951PLjSc
ajVba6dqlDi+q1H5tPvI5SWMN7PCBWIJ41+WvF+5IAZzQZYgNaVLbAAAAAAA
AAAEAAAAHwNiIi5Ms+4PsY/L2ipkTtquknfx6HodLvk3RAAAAAA==.
</pre></blockquote>

    <p>We could fragment this message into (for example) three
    pieces:</p>

<blockquote><pre>
?OTR|5a73a599|27e31597,00001,00003,?OTR:AAMDJ+MVmSfjFZcAAAAA
AQAAAAIAAADA1g5IjD1ZGLDVQEyCgCyn9hbrL3KAbGDdzE2ZkMyTKl7XfkSx
h8YJnudstiB74i4BzT0W2haClg6dMary/jo9sMudwmUdlnKpIGEKXWdvJKT+
hQ26h9nzMgEditLB8v,
</pre></blockquote>

<blockquote><pre>
?OTR|5a73a599|27e31597,00002,00003,jPEWAJ6gBXvZrY6ZQrx3gb4v0
UaSMOMiR5sB7Eaulb2Yc6RmRnnlxgUUC2alosg4WIeFN951PLjScajVba6dq
lDi+q1H5tPvI5SWMN7PCBWIJ41+WvF+5IAZzQZYgNaVLbAAAAAAAAAAEAAAA
HwNiIi5Ms+4PsY/L2i,
</pre></blockquote>

<blockquote><pre>
?OTR|5a73a599|27e31597,00003,00003,pkTtquknfx6HodLvk3RAAAAAA
==.,
</pre></blockquote>
<h3>The protocol state machine</h3>
<p>An OTR client maintains separate state for every correspondent.  For
example, Alice may have an active OTR conversation with Bob, while
having an unprotected conversation with Charlie.  This state consists of
two main state variables, as well as some other information (such as
encryption keys).  The two main state variables are:</p>
<h4>Message state</h4>
<p>The message state variable, msgstate, controls what happens to
outgoing messages typed by the user.  It can take one of three
values:</p>
<dl>
<dt>MSGSTATE_PLAINTEXT</dt>
<dd>This state indicates that outgoing messages are sent without
encryption.  This is the state that is used before an OTR conversation
is initiated.  This is the initial state, and the only way to
subsequently enter this state is for the user to explicitly request to
do so via some UI operation.</dd>
<dt>MSGSTATE_ENCRYPTED</dt>
<dd>This state indicates that outgoing messages are sent encrypted.
This is the state that is used during an OTR conversation.  The only way
to enter this state is for the authentication state machine (below) to
successfully complete.</dd>
<dt>MSGSTATE_FINISHED</dt>
<dd>This state indicates that outgoing messages are not delivered at
all.  This state is entered only when the other party indicates he has
terminated his side of the OTR conversation.  For example, if Alice and
Bob are having an OTR conversation, and Bob instructs his OTR client to
end its private session with Alice (for example, by logging out), Alice
will be notified of this, and <em>her</em> client will switch to
MSGSTATE_FINISHED mode.  This prevents Alice from accidentally sending a
message to Bob in plaintext.  (Consider what happens if Alice was in the
middle of typing a private message to Bob when he suddenly logs out,
just as Alice hits Enter.)</dd>
</dl>
<h4>Authentication state</h4>
<p>The authentication state variable, authstate, can take one of four
values (plus one extra for OTR version 1 compatibility):</p>
<dl>
<dt>AUTHSTATE_NONE</dt>
<dd>This state indicates that the authentication protocol is not
currently in progress.  This is the initial state.</dd>
<dt>AUTHSTATE_AWAITING_DHKEY</dt>
<dd>After Bob initiates the authentication protocol by sending Alice
the D-H Commit Message, he enters this state to await Alice's reply.</dd>
<dt>AUTHSTATE_AWAITING_REVEALSIG</dt>
<dd>After Alice receives Bob's D-H Commit Message, and replies with her
own D-H Key Message, she enters this state to await Bob's reply.</dd>
<dt>AUTHSTATE_AWAITING_SIG</dt>
<dd>After Bob receives Alice's D-H Key Message, and replies with his own
Reveal Signature Message, he enters this state to await Alice's reply.</dd>
<dt>AUTHSTATE_V1_SETUP</dt>
<dd>For OTR version 1 compatibility, if Bob sends a version 1 Key
Exchange Message to Alice, he enters this state to await Alice's
reply.</dd>
</dl>
<p>After:</p>
<ul>
<li>Alice (in AUTHSTATE_AWAITING_REVEALSIG) receives Bob's Reveal
Signature Message (and replies with her own Signature Message), <b>or</b>
</li>
<li>Bob (in AUTHSTATE_AWAITING_SIG) receives Alice's Signature Message,
/li>
</ul>
<p>then,
assuming the signature verifications succeed, the msgstate
variable is transitioned to MSGSTATE_ENCRYPTED.  Regardless of whether
the signature verifications succeed, the authstate variable is
transitioned to AUTHSTATE_NONE.</p>
<h4>Policies</h4>
<p>OTR clients can set different <b>policies</b> for different
correspondents.  For example, Alice could set up her client so that it
speaks only OTR version 3, except with Charlie, who she knows has only
an old client; so that it will opportunistically start an OTR conversation
whenever it detects the correspondent supports it; or so that it refuses
to send non-encrypted messages to Bob, ever.</p>
<p>The policies that can be set (on a global or per-correspondent basis)
are any combination of the following boolean flags:</p>
<dl>
<dt>ALLOW_V1</dt>
<dd>Allow version 1 of the OTR protocol to be used (in general this
document will not address OTR protocol version 1; see previous
protocol documents for these details).</dd>
<dt>ALLOW_V2</dt>
<dd>Allow version 2 of the OTR protocol to be used.</dd>
<dt>ALLOW_V3</dt>
<dd>Allow version 3 of the OTR protocol to be used.</dd>
<dt>REQUIRE_ENCRYPTION</dt>
<dd>Refuse to send unencrypted messages.</dd>
<dt>SEND_WHITESPACE_TAG</dt>
<dd>Advertise your support of OTR using the whitespace tag.</dd>
<dt>WHITESPACE_START_AKE</dt>
<dd>Start the OTR AKE when you receive a whitespace tag.</dd>
<dt>ERROR_START_AKE</dt>
<dd>Start the OTR AKE when you receive an OTR Error Message.</dd>
</dl>
<p>Note that it is possible for UIs simply to offer the old
"combinations" of options, and not ask about each one separately.</p>
<h4>State transitions</h4>
<p>There are twelve actions an OTR client must handle:</p>
<ul>
<li>Received messages:
<ul>
<li>Plaintext without the whitespace tag</li>
<li>Plaintext with the whitespace tag</li>
<li>Query Message</li>
<li>Error Message</li>
<li>D-H Commit Message</li>
<li>D-H Key Message</li>
<li>Reveal Signature Message</li>
<li>Signature Message</li>
<li>Data Message</li>
</ul></li>
<li>User actions:
<ul>
<li>User requests to start an OTR conversation</li>
<li>User requests to end an OTR conversation</li>
<li>User types a message to be sent</li>
</ul></li>
</ul>
<p>The following sections will outline what actions to take in each
case.  They all assume that at least one of ALLOW_V1, ALLOW_V2 or
ALLOW_V3 is set; if not, then OTR is completely disabled, and no
special handling of messages should be done at all. For version 1
messages, please refer to previous OTR protocol documents. For version
3 messages, someone receiving a message with a recipient instance tag
specified that does not equal their own should discard the message
and optionally warn the user. The exception here is the D-H Commit
Message where the recipient instance tag may be 0, indicating that no
particular instance is specified.</p>
<h4>Receiving plaintext without the whitespace tag</h4>
<dl>
<dt>If msgstate is MSGSTATE_PLAINTEXT:</dt>
<dd>Simply display the message to the user.  If REQUIRE_ENCRYPTION is
set, warn him that the message was received unencrypted.</dd>
<dt>If msgstate is MSGSTATE_ENCRYPTED or MSGSTATE_FINISHED:</dt>
<dd>Display the message to the user, but warn him that the message was
received unencrypted.</dd>
</dl>
<h4>Receiving plaintext with the whitespace tag</h4>
<dl>
<dt>If msgstate is MSGSTATE_PLAINTEXT:</dt>
<dd>Remove the whitespace tag and display the message to the user.  If
REQUIRE_ENCRYPTION is set, warn him that the message was received
unencrypted.</dd>
<dt>If msgstate is MSGSTATE_ENCRYPTED or MSGSTATE_FINISHED:</dt>
<dd>Remove the whitespace tag and display the message to the user, but
warn him that the message was received unencrypted.</dd>
</dl>
<p>In any event, if WHITESPACE_START_AKE is set:</p>
<dl>
<dt>If the tag offers OTR version 3 and ALLOW_V3 is set:</dt>
<dd>Send a version 3 D-H Commit Message, and transition authstate to
AUTHSTATE_AWAITING_DHKEY.</dd>
<dt>Otherwise, if the tag offers OTR version 2 and ALLOW_V2 is set:</dt>
<dd>Send a version 2 D-H Commit Message, and transition authstate to
AUTHSTATE_AWAITING_DHKEY.</dd>
</dl>
<h4>Receiving a Query Message</h4>
<dl>
<dt>If the query message offers OTR version 3 and ALLOW_V3 is set:</dt>
<dd>Send a version 3 D-H Commit Message, and transition authstate to
AUTHSTATE_AWAITING_DHKEY.</dd>
<dt>Otherwise, if the message offers OTR version 2 and ALLOW_V2 is set:</dt>
<dd>Send a version 2 D-H Commit Message, and transition authstate to
AUTHSTATE_AWAITING_DHKEY.</dd>
</dl>
<h4>Receiving an Error Message</h4>
<p>Display the message to the user.  If ERROR_START_AKE is set, reply
with a Query Message.</p>
<h4>User requests to start an OTR conversation</h4>
<p>Send an OTR Query Message to the correspondent.</p>
<h4>Receiving a D-H Commit Message</h4>
<p>If the message is version 2 and ALLOW_V2 is not set, ignore this message.
Similarly if the message is version 3 and ALLOW_V3 is not set, ignore the
message. Otherwise:</p>
<dl>
<dt>If authstate is AUTHSTATE_NONE:</dt>
<dd>Reply with a D-H Key Message, and transition authstate to
AUTHSTATE_AWAITING_REVEALSIG.</dd>
<dt>If authstate is AUTHSTATE_AWAITING_DHKEY:</dt>
<dd>This is the trickiest transition in the whole protocol.  It
indicates that you have already sent a D-H Commit message to your
correspondent, but that he either didn't receive it, or just didn't
receive it <em>yet</em>, and has sent you one as well.  The symmetry
will be broken by comparing the hashed g<sup>x</sup> you sent in your
D-H Commit Message with the one you received, considered as 32-byte
unsigned big-endian values.
<dl>
<dt>If yours is the higher hash value:</dt>
<dd>Ignore the incoming D-H Commit message, but resend your D-H
Commit message.</dd>
<dt>Otherwise:</dt>
<dd>Forget your old g<sup>x</sup> value that you sent (encrypted)
earlier, and pretend you're in AUTHSTATE_NONE; i.e. reply with a D-H Key
Message, and transition authstate to AUTHSTATE_AWAITING_REVEALSIG.</dd>
</dl></dd>
<dt>If authstate is AUTHSTATE_AWAITING_REVEALSIG:</dt>
<dd>Retransmit your D-H Key Message (the same
one as you sent when you entered AUTHSTATE_AWAITING_REVEALSIG).  Forget
the old D-H Commit message, and use this new one instead.  There
are a number of reasons this might happen, including:
<ul>
<li>Your correspondent simply started a new AKE.</li>
<li>Your correspondent resent his D-H Commit message, as specified
above.</li>
<li>On some networks, like AIM, if your correspondent is logged in
multiple times, each of his clients will send a D-H Commit Message in
response to a Query Message; resending the same D-H Key Message in
response to each of those messages will prevent compounded confusion,
since each of his clients will see each of the D-H Key Messages you
send.  [And the problem gets even worse if you are <em>each</em> logged
in multiple times.]</li>
</ul></dd>
<dt>If authstate is AUTHSTATE_AWAITING_SIG or AUTHSTATE_V1_SETUP:</dt>
<dd>Reply with a new D-H Key message, and transition authstate to
AUTHSTATE_AWAITING_REVEALSIG.</dd>
</dl>
<h4>Receiving a D-H Key Message</h4>
<p>If the message is version 2 and ALLOW_V2 is not set, ignore this
message. Similarly if the message is version 3 and ALLOW_V3 is not
set, ignore this message. Otherwise:</p>
<dl>
<dt>If authstate is AUTHSTATE_AWAITING_DHKEY:</dt>
<dd>Reply with a Reveal Signature Message and transition authstate to
AUTHSTATE_AWAITING_SIG.</dd>
<dt>If authstate is AUTHSTATE_AWAITING_SIG:</dt>
<dd>
<dl>
<dt>If this D-H Key message is the same the one you received earlier
(when you entered AUTHSTATE_AWAITING_SIG):</dt>
<dd>Retransmit your Reveal Signature Message.</dd>
<dt>Otherwise:</dt>
<dd>Ignore the message.</dd>
</dl></dd>
<dt>If authstate is AUTHSTATE_NONE, AUTHSTATE_AWAITING_REVEALSIG, or
AUTHSTATE_V1_SETUP:</dt>
<dd>Ignore the message.</dd>
</dl>
<h4>Receiving a Reveal Signature Message</h4>
<p>If the message is version 2 and ALLOW_V2 is not set, ignore this message.
Similarly if the message is version 3 and ALLOW_V3 is not set, ignore the
message. Otherwise:</p>
<dl>
<dt>If authstate is AUTHSTATE_AWAITING_REVEALSIG:</dt>
<dd>Use the received value of r to decrypt the value of g<sup>x</sup>
received in the D-H Commit Message, and verify the hash therein.
Decrypt the encrypted signature, and verify the signature and the MACs.
If everything checks out:
<ul>
<li>Reply with a Signature Message.</li>
<li>Transition authstate to AUTHSTATE_NONE.</li>
<li>Transition msgstate to MSGSTATE_ENCRYPTED.</li>
<li>If there is a recent stored message, encrypt it and send it as a
Data Message.</li>
</ul>
Otherwise, ignore the message.</dd>
<dt>If authstate is AUTHSTATE_NONE, AUTHSTATE_AWAITING_DHKEY, 
AUTHSTATE_AWAITING_SIG, or AUTHSTATE_V1_SETUP:</dt>
<dd>Ignore the message.</dd>
</dl>
<h4>Receiving a Signature Message</h4>
<p>If the message is version 2 and ALLOW_V2 is not set, ignore this message.
Similarly if the message is version 3 and ALLOW_V3 is not set, ignore the
message. Otherwise:</p>
<dl>
<dt>If authstate is AUTHSTATE_AWAITING_SIG:</dt>
<dd>Decrypt the encrypted signature, and verify the signature and the MACs.
If everything checks out:
<ul>
<li>Transition authstate to AUTHSTATE_NONE.</li>
<li>Transition msgstate to MSGSTATE_ENCRYPTED.</li>
<li>If there is a recent stored message, encrypt it and send it as a
Data Message.</li>
</ul>
Otherwise, ignore the message.</dd>
<dt>If authstate is AUTHSTATE_NONE, AUTHSTATE_AWAITING_DHKEY, 
or AUTHSTATE_AWAITING_REVEALSIG:</dt>
<dd>Ignore the message.</dd>
</dl>
<h4>User types a message to be sent</h4>
<dl>
<dt>If msgstate is MSGSTATE_PLAINTEXT:</dt>
<dd><dl><dt>If REQUIRE_ENCRYPTION is set:</dt>
<dd>Store the plaintext message for possible retransmission, and send a
Query Message.</dd>
<dt>Otherwise:</dt>
<dd>If SEND_WHITESPACE_TAG is set, and you have not received a plaintext
message from this correspondent since last entering MSGSTATE_PLAINTEXT,
attach the whitespace tag to the message.  Send the (possibly modified)
message as plaintext.</dd></dl></dd>
<dt>If msgstate is MSGSTATE_ENCRYPTED:</dt>
<dd>Encrypt the message, and send it as a Data Message.  Store the
plaintext message for possible retransmission.</dd>
<dt>If msgstate is MSGSTATE_FINISHED:</dt>
<dd>Inform the user that the message cannot be sent at this time.  Store
the plaintext message for possible retransmission.</dd>
</dl>
<h4>Receiving a Data Message</h4>
<dl>
<dt>If msgstate is MSGSTATE_ENCRYPTED:</dt>
<dd>Verify the information (MAC, keyids, ctr value, etc.) in the
message.
<dl>
<dt>If the verification succeeds:</dt>
<dd>
<ul>
<li>Decrypt the message and display the human-readable part (if
non-empty) to the user.</li>
<li>Update the D-H encryption keys, if necessary.</li>
<li>If you have not sent a message to this correspondent in some
(configurable) time, send a "heartbeat" message, consisting of a Data
Message encoding an empty plaintext.  The heartbeat message should have
the IGNORE_UNREADABLE flag set.</li>
<li>If the received message contains a TLV type 1, forget all encryption
keys for this correspondent, and transition msgstate to
MSGSTATE_FINISHED.</li>
</ul>
</dd>
<dt>Otherwise, inform the user that an unreadable encrypted message was
received, and reply with an Error Message.</dt>
</dl></dd>
<dt>If msgstate is MSGSTATE_PLAINTEXT or MSGSTATE_FINISHED:</dt>
<dd>Inform the user that an unreadable encrypted message was received,
and reply with an Error Message.</dd>
</dl>
<h4>User requests to end an OTR conversation</h4>
<dl>
<dt>If msgstate is MSGSTATE_PLAINTEXT:</dt>
<dd>Do nothing.</dd>
<dt>If msgstate is MSGSTATE_ENCRYPTED:</dt>
<dd>Send a Data Message, encoding a message with an empty human-readable
part, and TLV type 1.  Transition msgstate to MSGSTATE_PLAINTEXT.</dd>
<dt>If msgstate is MSGSTATE_FINISHED:</dt>
<dd>Transition msgstate to MSGSTATE_PLAINTEXT.</dd>
</dl>
</body></html>