tor-spec.txt 22 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523
  1. $Id$
  2. Tor Spec
  3. Note: This is an attempt to specify Tor as it exists as implemented in
  4. early June, 2003. It is not recommended that others implement this
  5. design as it stands; future versions of Tor will implement improved
  6. protocols.
  7. TODO: (very soon)
  8. - Specify truncate/truncated payloads?
  9. - Specify RELAY_END payloads. [It's 1 byte of reason, then X bytes of
  10. data, right?]
  11. - Sendme w/stream0 is circuit sendme
  12. - Integrate -NM and -RD comments
  13. - EXTEND cells should have hostnames or nicknames, so that OPs never
  14. resolve OR hostnames. Else DNS servers can give different answers to
  15. different OPs, and compromise their anonymity.
  16. EVEN LATER:
  17. - Do TCP-style sequencing and ACKing of DATA cells so that we can afford
  18. to lose some data cells.
  19. 0. Notation:
  20. PK -- a public key.
  21. SK -- a private key
  22. K -- a key for a symmetric cypher
  23. a|b -- concatenation of 'a' with 'b'.
  24. All numeric values are encoded in network (big-endian) order.
  25. Unless otherwise specified, all symmetric ciphers are AES in counter
  26. mode, with an IV of all 0 bytes. Asymmetric ciphers are either RSA
  27. with 1024-bit keys and exponents of 65537, or DH with the safe prime
  28. from rfc2409, section 6.2, whose hex representation is:
  29. "FFFFFFFFFFFFFFFFC90FDAA22168C234C4C6628B80DC1CD129024E08"
  30. "8A67CC74020BBEA63B139B22514A08798E3404DDEF9519B3CD3A431B"
  31. "302B0A6DF25F14374FE1356D6D51C245E485B576625E7EC6F44C42E9"
  32. "A637ED6B0BFF5CB6F406B7EDEE386BFB5A899FA5AE9F24117C4B1FE6"
  33. "49286651ECE65381FFFFFFFFFFFFFFFF"
  34. 1. System overview
  35. Tor is a connection-oriented anonymizing communication service. Users
  36. build a path known as a "virtual circuit" through the network, in which
  37. each node knows its predecessor and successor, but no others. Traffic
  38. flowing down the circuit is unwrapped by a symmetric key at each node,
  39. which reveals the downstream node.
  40. 2. Connections
  41. There are two ways to connect to an onion router (OR). The first is
  42. as an onion proxy (OP), which allows the OP to authenticate the OR
  43. without authenticating itself. The second is as another OR, which
  44. allows mutual authentication.
  45. Tor uses TLS for link encryption, using the cipher suite
  46. "TLS_DHE_RSA_WITH_AES_128_CBC_SHA". An OR always sends a
  47. self-signed X.509 certificate whose commonName is the server's
  48. nickname, and whose public key is in the server directory.
  49. All parties receiving certificates must confirm that the public
  50. key is as it appears in the server directory, and close the
  51. connection if it is not.
  52. Once a TLS connection is established, the two sides send cells
  53. (specified below) to one another. Cells are sent serially. All
  54. cells are 256 bytes long. Cells may be sent embedded in TLS
  55. records of any size or divided across TLS records, but the framing
  56. of TLS records should not leak information about the type or
  57. contents of the cells.
  58. OR-to-OR connections are never deliberately closed. An OP should
  59. close a connection to an OR if there are no circuits running over
  60. the connection, and an amount of time (KeepalivePeriod, defaults to
  61. 5 minutes) has passed.
  62. 3. Cell Packet format
  63. The basic unit of communication for onion routers and onion
  64. proxies is a fixed-width "cell". Each cell contains the following
  65. fields:
  66. CircID [2 bytes]
  67. Command [1 byte]
  68. Length [1 byte]
  69. Sequence number (unused, set to 0) [4 bytes]
  70. Payload (padded with 0 bytes) [248 bytes]
  71. [Total size: 256 bytes]
  72. The 'Command' field holds one of the following values:
  73. 0 -- PADDING (Padding) (See Sec 6.2)
  74. 1 -- CREATE (Create a circuit) (See Sec 4)
  75. 2 -- CREATED (Acknowledge create) (See Sec 4)
  76. 3 -- RELAY (End-to-end data) (See Sec 5)
  77. 4 -- DESTROY (Stop using a circuit) (See Sec 4)
  78. The interpretation of 'Length' and 'Payload' depend on the type of
  79. the cell.
  80. PADDING: Neither field is used.
  81. CREATE: Length is 144; the payload contains the first phase of the
  82. DH handshake.
  83. CREATED: Length is 128; the payload contains the second phase of
  84. the DH handshake.
  85. RELAY: Length is a value between 8 and 248; the first 'length'
  86. bytes of payload contain useful data.
  87. DESTROY: Neither field is used.
  88. Unused fields are filled with 0 bytes. The payload is padded with
  89. 0 bytes.
  90. PADDING cells are currently used to implement connection
  91. keepalive. ORs and OPs send one another a PADDING cell every few
  92. minutes.
  93. CREATE and DESTROY cells are used to manage circuits; see section
  94. 4 below.
  95. RELAY cells are used to send commands and data along a circuit; see
  96. section 5 below.
  97. 4. Circuit management
  98. 4.1. CREATE and CREATED cells
  99. Users set up circuits incrementally, one hop at a time. To create
  100. a new circuit, users send a CREATE cell to the first node, with the
  101. first half of the DH handshake; that node responds with a CREATED cell
  102. with the second half of the DH handshake. To extend a circuit past
  103. the first hop, the user sends an EXTEND relay cell (see section 5)
  104. which instructs the last node in the circuit to send a CREATE cell
  105. to extend the circuit.
  106. The payload for a CREATE cell is an 'onion skin', consisting of:
  107. RSA-encrypted data [128 bytes]
  108. Symmetrically-encrypted data [16 bytes]
  109. The RSA-encrypted portion contains:
  110. Symmetric key [16 bytes]
  111. First part of DH data (g^x) [112 bytes]
  112. The symmetrically encrypted portion contains:
  113. Second part of DH data (g^x) [16 bytes]
  114. The two parts of the DH data, once decrypted and concatenated, form
  115. g^x as calculated by the client.
  116. The relay payload for an EXTEND relay cell consists of:
  117. Address [4 bytes]
  118. Port [2 bytes]
  119. Onion skin [144 bytes]
  120. The port and address field denote the IPV4 address and port of the
  121. next onion router in the circuit.
  122. 4.2. Setting circuit keys
  123. Once the handshake between the OP and an OR is completed, both
  124. servers can now calculate g^xy with ordinary DH. From the base key
  125. material g^xy, they compute two 16 byte keys, called Kf and Kb as
  126. follows. First, the server represents g^xy as a big-endian
  127. unsigned integer. Next, the server computes 40 bytes of key data
  128. as K = SHA1(g^xy | [00]) | SHA1(g^xy | [01]) where "00" is a single
  129. octet whose value is zero, and "01" is a single octet whose value
  130. is one. The first 16 bytes of K form Kf, and the next 16 bytes of
  131. K form Kb.
  132. Kf is used to encrypt the stream of data going from the OP to the
  133. OR, whereas Kb is used to encrypt the stream of data going from the
  134. OR to the OP.
  135. 4.3. Creating circuits
  136. When creating a circuit through the network, the circuit creator
  137. performs the following steps:
  138. 1. Choose a chain of N onion routers (R_1...R_N) to constitute
  139. the path, such that no router appears in the path twice.
  140. [this is wrong, see October 2003 discussion on or-dev]
  141. 2. If not already connected to the first router in the chain,
  142. open a new connection to that router.
  143. 3. Choose a circID not already in use on the connection with the
  144. first router in the chain. If we are an onion router and our
  145. nickname is lexicographically greater than the nickname of the
  146. other side, then let the high bit of the circID be 1, else 0.
  147. 4. Send a CREATE cell along the connection, to be received by
  148. the first onion router.
  149. 5. Wait until a CREATED cell is received; finish the handshake
  150. and extract the forward key Kf_1 and the back key Kb_1.
  151. 6. For each subsequent onion router R (R_2 through R_N), extend
  152. the circuit to R.
  153. To extend the circuit by a single onion router R_M, the circuit
  154. creator performs these steps:
  155. 1. Create an onion skin, encrypting the RSA-encrypted part with
  156. R's public key.
  157. 2. Encrypt and send the onion skin in a relay EXTEND cell along
  158. the circuit (see section 5).
  159. 3. When a relay EXTENDED cell is received, calculate the shared
  160. keys. The circuit is now extended.
  161. When an onion router receives an EXTEND relay cell, it sends a
  162. CREATE cell to the next onion router, with the enclosed onion skin
  163. as its payload. The initiating onion router chooses some circID not
  164. yet used on the connection between the two onion routers. (But see
  165. section 4.3. above, concerning choosing circIDs.)
  166. As an extension (called router twins), if the desired next onion
  167. router R in the circuit is down, and some other onion router R'
  168. has the same key as R, then it's ok to extend to R' rather than R.
  169. When an onion router receives a CREATE cell, if it already has a
  170. circuit on the given connection with the given circID, it drops the
  171. cell. Otherwise, sometime after receiving the CREATE cell, it completes
  172. the DH handshake, and replies with a CREATED cell, containing g^y
  173. as its [128 byte] payload. Upon receiving a CREATED cell, an onion
  174. router packs it payload into an EXTENDED relay cell (see section 5),
  175. and sends that cell up the circuit. Upon receiving the EXTENDED
  176. relay cell, the OP can retrieve g^y.
  177. (As an optimization, OR implementations may delay processing onions
  178. until a break in traffic allows time to do so without harming
  179. network latency too greatly.)
  180. 4.4. Tearing down circuits
  181. Circuits are torn down when an unrecoverable error occurs along
  182. the circuit, or when all streams on a circuit are closed and the
  183. circuit's intended lifetime is over. Circuits may be torn down
  184. either completely or hop-by-hop.
  185. To tear down a circuit completely, an OR or OP sends a DESTROY
  186. cell to the adjacent nodes on that circuit, using the appropriate
  187. direction's circID.
  188. Upon receiving an outgoing DESTROY cell, an OR frees resources
  189. associated with the corresponding circuit. If it's not the end of
  190. the circuit, it sends a DESTROY cell for that circuit to the next OR
  191. in the circuit. If the node is the end of the circuit, then it tears
  192. down any associated edge connections (see section 5.1).
  193. After a DESTROY cell has been processed, an OR ignores all data or
  194. destroy cells for the corresponding circuit.
  195. To tear down part of a circuit, the OP sends a RELAY_TRUNCATE cell
  196. signaling a given OR (Stream ID zero). That OR sends a DESTROY
  197. cell to the next node in the circuit, and replies to the OP with a
  198. RELAY_TRUNCATED cell.
  199. When an unrecoverable error occurs along one connection in a
  200. circuit, the nodes on either side of the connection should, if they
  201. are able, act as follows: the node closer to the OP should send a
  202. RELAY_TRUNCATED cell towards the OP; the node farther from the OP
  203. should send a DESTROY cell down the circuit.
  204. [We'll have to reevaluate this section once we figure out cleaner
  205. circuit/connection killing conventions. -RD]
  206. 4.5. Routing data cells
  207. When an OR receives a RELAY cell, it checks the cell's circID and
  208. determines whether it has a corresponding circuit along that
  209. connection. If not, the OR drops the RELAY cell.
  210. Otherwise, if the OR is not at the OP edge of the circuit (that is,
  211. either an 'exit node' or a non-edge node), it de/encrypts the length
  212. field and the payload with AES/CTR, as follows:
  213. 'Forward' relay cell (same direction as CREATE):
  214. Use Kf as key; encrypt.
  215. 'Back' relay cell (opposite direction from CREATE):
  216. Use Kb as key; decrypt.
  217. If the OR recognizes the stream ID on the cell (it is either the ID
  218. of an open stream or the signaling (zero) ID), the OR processes the
  219. contents of the relay cell. Otherwise, it passes the decrypted
  220. relay cell along the circuit if the circuit continues, or drops the
  221. cell if it's the end of the circuit. [Getting an unrecognized
  222. relay cell at the end of the circuit must be allowed for now;
  223. we can reexamine this once we've designed full tcp-style close
  224. handshakes. -RD]
  225. Otherwise, if the data cell is coming from the OP edge of the
  226. circuit, the OP decrypts the length and payload fields with AES/CTR as
  227. follows:
  228. OP sends data cell to node R_M:
  229. For I=1...M, decrypt with Kf_I.
  230. Otherwise, if the data cell is arriving at the OP edge if the
  231. circuit, the OP encrypts the length and payload fields with AES/CTR as
  232. follows:
  233. OP receives data cell:
  234. For I=N...1,
  235. Encrypt with Kb_I. If the stream ID is a recognized
  236. stream for R_I, or if the stream ID is the signaling
  237. ID (zero), then stop and process the payload.
  238. For more information, see section 5 below.
  239. 5. Application connections and stream management
  240. 5.1. Streams
  241. Within a circuit, the OP and the exit node use the contents of
  242. RELAY packets to tunnel end-to-end commands and TCP connections
  243. ("Streams") across circuits. End-to-end commands can be initiated
  244. by either edge; streams are initiated by the OP.
  245. The first 8 bytes of each relay cell are reserved as follows:
  246. Relay command [1 byte]
  247. Stream ID [7 bytes]
  248. The relay commands are:
  249. 1 -- RELAY_BEGIN
  250. 2 -- RELAY_DATA
  251. 3 -- RELAY_END
  252. 4 -- RELAY_CONNECTED
  253. 5 -- RELAY_SENDME
  254. 6 -- RELAY_EXTEND
  255. 7 -- RELAY_EXTENDED
  256. 8 -- RELAY_TRUNCATE
  257. 9 -- RELAY_TRUNCATED
  258. 10 -- RELAY_DROP
  259. All RELAY cells pertaining to the same tunneled stream have the
  260. same stream ID. Stream ID's are chosen randomly by the OP. A
  261. stream ID is considered "recognized" on a circuit C by an OP or an
  262. OR if it already has an existing stream established on that
  263. circuit, or if the stream ID is equal to the signaling stream ID,
  264. which is all zero: [00 00 00 00 00 00 00]
  265. To create a new anonymized TCP connection, the OP sends a
  266. RELAY_BEGIN data cell with a payload encoding the address and port
  267. of the destination host. The stream ID is zero. The payload format is:
  268. NEWSTREAMID | ADDRESS | ':' | PORT | '\000'
  269. where NEWSTREAMID is the newly generated Stream ID to use for
  270. this stream, ADDRESS may be a DNS hostname, or an IPv4 address in
  271. dotted-quad format; and where PORT is encoded in decimal.
  272. Upon receiving this packet, the exit node resolves the address as
  273. necessary, and opens a new TCP connection to the target port. If
  274. the address cannot be resolved, or a connection can't be
  275. established, the exit node replies with a RELAY_END cell.
  276. Otherwise, the exit node replies with a RELAY_CONNECTED cell.
  277. The OP waits for a RELAY_CONNECTED cell before sending any data.
  278. Once a connection has been established, the OP and exit node
  279. package stream data in RELAY_DATA cells, and upon receiving such
  280. cells, echo their contents to the corresponding TCP stream.
  281. Relay RELAY_DROP cells are long-range dummies; upon receiving such
  282. a cell, the OR or OP must drop it.
  283. 5.2. Closing streams
  284. [Note -- TCP streams can only be half-closed for reading. Our
  285. Bickford's conversation was incorrect. -NM]
  286. Because TCP connections can be half-open, we follow an equivalent
  287. to TCP's FIN/FIN-ACK/ACK protocol to close streams.
  288. An exit connection can have a TCP stream in one of three states:
  289. 'OPEN', 'DONE_PACKAGING', and 'DONE_DELIVERING'. For the purposes
  290. of modeling transitions, we treat 'CLOSED' as a fourth state,
  291. although connections in this state are not, in fact, tracked by the
  292. onion router.
  293. A stream begins in the 'OPEN' state. Upon receiving a 'FIN' from
  294. the corresponding TCP connection, the edge node sends a 'RELAY_END'
  295. cell along the circuit and changes its state to 'DONE_PACKAGING'.
  296. Upon receiving a 'RELAY_END' cell, an edge node sends a 'FIN' to
  297. the corresponding TCP connection (e.g., by calling
  298. shutdown(SHUT_WR)) and changing its state to 'DONE_DELIVERING'.
  299. When a stream in already in 'DONE_DELIVERING' receives a 'FIN', it
  300. also sends a 'RELAY_END' along the circuit, and changes its state
  301. to 'CLOSED'. When a stream already in 'DONE_PACKAGING' receives a
  302. 'RELAY_END' cell, it sends a 'FIN' and changes its state to
  303. 'CLOSED'.
  304. [Note: Please rename 'RELAY_END2'. :) -NM ]
  305. If an edge node encounters an error on any stram, it sends a
  306. 'RELAY_END2' cell along the circuit (if possible) and closes the
  307. TCP connection immediately. If an edge node receives a
  308. 'RELAY_END2' cell for any stream, it closes the TCP connection
  309. completely, and sends nothing along the circuit.
  310. 6. Flow control
  311. 6.1. Link throttling
  312. Each node should do appropriate bandwidth throttling to keep its
  313. user happy.
  314. Communicants rely on TCP's default flow control to push back when they
  315. stop reading.
  316. 6.2. Link padding
  317. Currently nodes are not required to do any sort of link padding or
  318. dummy traffic. Because strong attacks exist even with link padding,
  319. and because link padding greatly increases the bandwidth requirements
  320. for running a node, we plan to leave out link padding until this
  321. tradeoff is better understood.
  322. 6.3. Circuit-level flow control
  323. To control a circuit's bandwidth usage, each OR keeps track of
  324. two 'windows', consisting of how many RELAY_DATA cells it is
  325. allowed to package for transmission, and how many RELAY_DATA cells
  326. it is willing to deliver to streams outside the network.
  327. Each 'window' value is initially set to 1000 data cells
  328. in each direction (cells that are not data cells do not affect
  329. the window). When an OR is willing to deliver more cells, it sends a
  330. RELAY_SENDME cell towards the OP, with Stream ID zero. When an OR
  331. receives a RELAY_SENDME cell with stream ID zero, it increments its
  332. packaging window.
  333. Each of these cells increments the corresponding window by 100.
  334. The OP behaves identically, except that it must track a packaging
  335. window and a delivery window for every OR in the circuit.
  336. An OR or OP sends cells to increment its delivery window when the
  337. corresponding window value falls under some threshold (900).
  338. If a packaging window reaches 0, the OR or OP stops reading from
  339. TCP connections for all streams on the corresponding circuit, and
  340. sends no more RELAY_DATA cells until receiving a RELAY_SENDME cell.
  341. [this stuff is badly worded; copy in the tor-design section -RD]
  342. 6.4. Stream-level flow control
  343. Edge nodes use RELAY_SENDME cells to implement end-to-end flow
  344. control for individual connections across circuits. Similarly to
  345. circuit-level flow control, edge nodes begin with a window of cells
  346. (500) per stream, and increment the window by a fixed value (50)
  347. upon receiving a RELAY_SENDME cell. Edge nodes initiate RELAY_SENDME
  348. cells when both a) the window is <= 450, and b) there are less than
  349. ten cell payloads remaining to be flushed at that edge.
  350. 7. Directories and routers
  351. 7.1. Router descriptor format.
  352. (Unless otherwise noted, tokens on the same line are space-separated.)
  353. Router ::= Router-Line Date-Line Onion-Key Link-Key Signing-Key Exit-Policy Router-Signature NL
  354. Router-Line ::= "router" nickname address ORPort SocksPort DirPort bandwidth NL
  355. Date-Line ::= "published" YYYY-MM-DD HH:MM:SS NL
  356. Onion-key ::= "onion-key" NL a public key in PEM format NL
  357. Link-key ::= "link-key" NL a public key in PEM format NL
  358. Signing-Key ::= "signing-key" NL a public key in PEM format NL
  359. Exit-Policy ::= Exit-Line*
  360. Exit-Line ::= ("accept"|"reject") string NL
  361. Router-Signature ::= "router-signature" NL Signature
  362. Signature ::= "-----BEGIN SIGNATURE-----" NL
  363. Base-64-encoded-signature NL "-----END SIGNATURE-----" NL
  364. ORport ::= port where the router listens for routers/proxies (speaking cells)
  365. SocksPort ::= where the router listens for applications (speaking socks)
  366. DirPort ::= where the router listens for directory download requests
  367. bandwidth ::= maximum bandwidth, in bytes/s
  368. nickname ::= between 1 and 32 alphanumeric characters. case-insensitive.
  369. Example:
  370. router moria1 moria.mit.edu 9001 9021 9031 100000
  371. published 2003-09-24 19:36:05
  372. -----BEGIN RSA PUBLIC KEY-----
  373. MIGJAoGBAMBBuk1sYxEg5jLAJy86U3GGJ7EGMSV7yoA6mmcsEVU3pwTUrpbpCmwS
  374. 7BvovoY3z4zk63NZVBErgKQUDkn3pp8n83xZgEf4GI27gdWIIwaBjEimuJlEY+7K
  375. nZ7kVMRoiXCbjL6VAtNa4Zy1Af/GOm0iCIDpholeujQ95xew7rQnAgMA//8=
  376. -----END RSA PUBLIC KEY-----
  377. signing-key
  378. -----BEGIN RSA PUBLIC KEY-----
  379. 7BvovoY3z4zk63NZVBErgKQUDkn3pp8n83xZgEf4GI27gdWIIwaBjEimuJlEY+7K
  380. MIGJAoGBAMBBuk1sYxEg5jLAJy86U3GGJ7EGMSV7yoA6mmcsEVU3pwTUrpbpCmwS
  381. f/GOm0iCIDpholeujQ95xew7rnZ7kVMRoiXCbjL6VAtNa4Zy1AQnAgMA//8=
  382. -----END RSA PUBLIC KEY-----
  383. reject 18.0.0.0/24
  384. Note: The extra newline at the end of the router block is intentional.
  385. 7.2. Directory format
  386. Directory ::= Directory-Header Directory-Router Router* Signature
  387. Directory-Header ::= "signed-directory" NL Software-Line NL
  388. Software-Line: "recommended-software" comma-separated-version-list
  389. Directory-Router ::= Router
  390. Directory-Signature ::= "directory-signature" NL Signature
  391. Signature ::= "-----BEGIN SIGNATURE-----" NL
  392. Base-64-encoded-signature NL "-----END SIGNATURE-----" NL
  393. Note: The router block for the directory server must appear first.
  394. The signature is computed by computing the SHA-1 hash of the
  395. directory, from the characters "signed-directory", through the newline
  396. after "directory-signature". This digest is then padded with PKCS.1,
  397. and signed with the directory server's signing key.
  398. 7.3. Behavior of a directory server
  399. lists nodes that are connected currently
  400. speaks http on a socket, spits out directory on request
  401. -----------
  402. (for emacs)
  403. Local Variables:
  404. mode:text
  405. indent-tabs-mode:nil
  406. fill-column:77
  407. End: