## translation metadata
 # Revision: $Revision$
 # Translation-Priority: 3-low
 #include "head.wmi" TITLE="Tor: Onion Service Protocol" CHARSET="UTF-8"
 <div id="content" class="clearfix">
   <div id="breadcrumbs">
     <a href="<page index>">Home &raquo; </a>
     <a href="<page docs/documentation>">Documentation &raquo; </a>
     <a href="<page docs/onion-services>">Onion Services</a>
   <div id="maincol">
     <h2>Tor: Onion Service Protocol</h2>
     Tor makes it possible for users to hide their locations while offering
     various kinds of services, such as web publishing or an instant
     messaging server.  Using Tor "rendezvous points," other Tor users can
 	connect to these onion services, formerly known as hidden services, each
 	without knowing the other's network identity. This page describes the
 	technical details of how this rendezvous protocol works. For a more direct
 	how-to, see our <a href="<page docs/tor-onion-service>">configuring onion
 	services</a> page.  </p>
     An onion service needs to advertise its existence in the Tor network before
     clients will be able to contact it. Therefore, the service randomly picks
     some relays, builds circuits to them, and asks them to act as
     <em>introduction points</em> by telling them its public key. Note
     that in the following figures the green links are circuits rather
     than direct connections. By using a full Tor circuit, it's hard for
     anyone to associate an introduction point with the onion server's IP
     address. While the introduction points and others are told the onion
     service's identity (public key), we don't want them to learn about the
     onion server's location (IP address).
     <img alt="Tor onion service step one" src="$(IMGROOT)/tor-onion-services-1.png">
     # maybe add a speech bubble containing "PK" to Bob, because that's what
     # Bob tells to his introduction points
 	Step two: the onion service assembles an <em>onion service descriptor</em>,
 	containing its public key and a summary of each introduction point, and
 	signs this descriptor with its private key.  It uploads that descriptor to
 	a distributed hash table.  The descriptor will be found by clients
 	requesting XYZ.onion where XYZ is a 16 character name derived from the
 	service's public key. After this step, the onion service is set up.  </p>
     Although it might seem impractical to use an automatically-generated
     service name, it serves an important goal: Everyone &ndash; including
 	the introduction points, the distributed hash table directory, and of
 	course the clients &ndash; can verify that they are talking to the right
 	onion service. See also <a
 	conjecture</a> that out of Decentralized, Secure, and Human-Meaningful, you
 	can achieve at most two. Perhaps one day somebody will implement a <a
 	design for onion service names?  </p>
     <img alt="Tor onion service step two" src="$(IMGROOT)/tor-onion-services-2.png">
     # maybe replace "database" with "DHT"; further: how incorrect
     # is it to *not* add DB to the Tor cloud, now that begin dir cells are in
     # use?
     Step three: A client that wants to contact an onion service needs
     to learn about its onion address first. After that, the client can
     initiate connection establishment by downloading the descriptor from
     the distributed hash table. If there is a descriptor for XYZ.onion
     (the onion service could also be offline or have left long ago,
     or there could be a typo in the onion address), the client now
     knows the set of introduction points and the right public key to
     use. Around this time, the client also creates a circuit to another
     randomly picked relay and asks it to act as <em>rendezvous point</em>
     by telling it a one-time secret.
     <img alt="Tor onion service step three" src="$(IMGROOT)/tor-onion-services-3.png">
     # maybe add "cookie" to speech bubble, separated from the surrounded
     # "IP1-3" and "PK"
     Step four: When the descriptor is present and the rendezvous
     point is ready, the client assembles an <em>introduce</em> message
     (encrypted to the onion service's public key) including the address
     of the rendezvous point and the one-time secret. The client sends
     this message to one of the introduction points, requesting it be
     delivered to the onion service. Again, communication takes place
     via a Tor circuit: nobody can relate sending the introduce message
     to the client's IP address, so the client remains anonymous.
     <img alt="Tor onion service step four" src="$(IMGROOT)/tor-onion-services-4.png">
     Step five: The onion service decrypts the client's introduce message
     and finds the address of the rendezvous point and the one-time secret
     in it. The service creates a circuit to the rendezvous point and
     sends the one-time secret to it in a rendezvous message.
     At this point it is of special importance that the onion service sticks to
     the same set of <a
     guards</a> when creating new circuits. Otherwise an attacker
     could run their own relay and force an onion service to create an arbitrary
     number of circuits in the hope that the corrupt relay is picked as entry
     node and they learn the onion server's IP address via timing analysis. This
     attack was described by &Oslash;verlier and Syverson in their paper titled
     <a href="">Locating Hidden
     <img alt="Tor onion service step five" src="$(IMGROOT)/tor-onion-services-5.png">
     # it should say "Bob connects to Alice's ..."
     In the last step, the rendezvous point notifies the client about successful
     connection establishment. After that, both client and onion service can
     use their circuits to the rendezvous point for communicating with each
     other. The rendezvous point simply relays (end-to-end encrypted) messages
     from client to service and vice versa.
     One of the reasons for not using the introduction circuit
     for actual communication is that no single relay should
     appear to be responsible for a given onion service. This is why the
     rendezvous point never learns about the onion service's identity.
     In general, the complete connection between client and onion service
     consists of 6 relays: 3 of them were picked by the client with the third
     being the rendezvous point and the other 3 were picked by the onion
     <img alt="Tor onion service step six" src="$(IMGROOT)/tor-onion-services-6.png">
     There are more detailed descriptions about the onion service protocol than
     this one. See the
     <a href="<svnprojects>design-paper/tor-design.pdf">Tor design paper</a>
     for an in-depth design description and the
     <a href="<specblob>rend-spec-v3.txt">rendezvous specification</a>
     for the message formats.
   <!-- END MAINCOL -->
   <div id = "sidecol">
 #include "side.wmi"
 #include "info.wmi"
   <!-- END SIDECOL -->
 <!-- END CONTENT -->
 #include <foot.wmi>