tor-spec.txt 21 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520
  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
  9. - Sendme w/stream0 is circuit sendme
  10. - Integrate -NM and -RD comments
  11. - EXTEND cells should have hostnames or nicknames, so that OPs never
  12. resolve OR hostnames. Else DNS servers can give different answers to
  13. different OPs, and compromise their anonymity.
  14. EVEN LATER:
  15. - Do TCP-style sequencing and ACKing of DATA cells so that we can afford
  16. to lose some data cells.
  17. -
  18. 0. Notation:
  19. PK -- a public key.
  20. SK -- a private key
  21. K -- a key for a symmetric cypher
  22. a|b -- concatenation of 'a' with 'b'.
  23. All numeric values are encoded in network (big-endian) order.
  24. Unless otherwise specified, all symmetric ciphers are AES in counter
  25. mode, with an IV of all 0 bytes. Asymmetric ciphers are either RSA
  26. with 1024-bit keys and exponents of 65537, or DH with the safe prime
  27. from rfc2409, section 6.2, whose hex representation is:
  28. "FFFFFFFFFFFFFFFFC90FDAA22168C234C4C6628B80DC1CD129024E08"
  29. "8A67CC74020BBEA63B139B22514A08798E3404DDEF9519B3CD3A431B"
  30. "302B0A6DF25F14374FE1356D6D51C245E485B576625E7EC6F44C42E9"
  31. "A637ED6B0BFF5CB6F406B7EDEE386BFB5A899FA5AE9F24117C4B1FE6"
  32. "49286651ECE65381FFFFFFFFFFFFFFFF"
  33. 1. System overview
  34. Tor is a connection-oriented anonymizing communication service. Users
  35. build a path known as a "virtual circuit" through the network, in which
  36. each node knows its predecessor and successor, but no others. Traffic
  37. flowing down the circuit is unwrapped by a symmetric key at each node,
  38. which reveals the downstream node.
  39. 2. Connections
  40. There are two ways to connect to an onion router (OR). The first is
  41. as an onion proxy (OP), which allows the OP to authenticate the OR
  42. without authenticating itself. The second is as another OR, which
  43. allows mutual authentication.
  44. Tor uses TLS for link encryption, using the cipher suite
  45. "TLS_DHE_RSA_WITH_AES_128_CBC_SHA". An OR always sends a
  46. self-signed X.509 certificate whose commonName is the server's
  47. nickname, and whose public key is in the server directory.
  48. All parties receiving certificates must confirm that the public
  49. key is as it appears in the server directory, and close the
  50. connection if it is not.
  51. Once a TLS connection is established, the two sides send cells
  52. (specified below) to one another. Cells are sent serially. All
  53. cells are 256 bytes long. Cells may be sent embedded in TLS
  54. records of any size or divided across TLS records, but the framing
  55. of TLS records should not leak information about the type or
  56. contents of the cells.
  57. OR-to-OR connections are never deliberately closed. An OP should
  58. close a connection to an OR if there are no circuits running over
  59. the connection, and an amount of time (KeepalivePeriod, defaults to
  60. 5 minutes) has passed.
  61. 3. Cell Packet format
  62. The basic unit of communication for onion routers and onion
  63. proxies is a fixed-width "cell". Each cell contains the following
  64. fields:
  65. ACI (anonymous circuit identifier) [2 bytes]
  66. Command [1 byte]
  67. Length [1 byte]
  68. Sequence number (unused, set to 0) [4 bytes]
  69. Payload (padded with 0 bytes) [248 bytes]
  70. [Total size: 256 bytes]
  71. The 'Command' field holds one of the following values:
  72. 0 -- PADDING (Padding) (See Sec 6.2)
  73. 1 -- CREATE (Create a circuit) (See Sec 4)
  74. 2 -- CREATED (Acknowledge create) (See Sec 4)
  75. 3 -- RELAY (End-to-end data) (See Sec 5)
  76. 4 -- DESTROY (Stop using a circuit) (See Sec 4)
  77. The interpretation of 'Length' and 'Payload' depend on the type of
  78. the cell.
  79. PADDING: Neither field is used.
  80. CREATE: Length is 144; the payload contains the first phase of the
  81. DH handshake.
  82. CREATED: Length is 128; the payload contains the second phase of
  83. the DH handshake.
  84. RELAY: Length is a value between 8 and 248; the first 'length'
  85. bytes of payload contain useful data.
  86. DESTROY: Neither field is used.
  87. Unused fields are filled with 0 bytes. The payload is padded with
  88. 0 bytes.
  89. PADDING cells are currently used to implement connection
  90. keepalive. ORs and OPs send one another a PADDING cell every few
  91. minutes.
  92. CREATE and DESTROY cells are used to manage circuits; see section
  93. 4 below.
  94. RELAY cells are used to send commands and data along a circuit; see
  95. section 5 below.
  96. 4. Circuit management
  97. 4.1. CREATE and CREATED cells
  98. Users set up circuits incrementally, one hop at a time. To create
  99. a new circuit, users send a CREATE cell to the first node, with the
  100. first half of the DH handshake; that node responds with a CREATED cell
  101. with the second half of the DH handshake. To extend a circuit past
  102. the first hop, the user sends an EXTEND relay cell (see section 5)
  103. which instructs the last node in the circuit to send a CREATE cell
  104. to extend the circuit.
  105. The payload for a CREATE cell is an 'onion skin', consisting of:
  106. RSA-encrypted data [128 bytes]
  107. Symmetrically-encrypted data [16 bytes]
  108. The RSA-encrypted portion contains:
  109. Symmetric key [16 bytes]
  110. First part of DH data (g^x) [112 bytes]
  111. The symmetrically encrypted portion contains:
  112. Second part of DH data (g^x) [16 bytes]
  113. The two parts of the DH data, once decrypted and concatenated, form
  114. g^x as calculated by the client.
  115. The relay payload for an EXTEND relay cell consists of:
  116. Address [4 bytes]
  117. Port [2 bytes]
  118. Onion skin [144 bytes]
  119. The port and address field denote the IPV4 address and port of the
  120. next onion router in the circuit.
  121. 4.2. Setting circuit keys
  122. Once the handshake between the OP and an OR is completed, both
  123. servers can now calculate g^xy with ordinary DH. From the base key
  124. material g^xy, they compute two 16 byte keys, called Kf and Kb as
  125. follows. First, the server represents g^xy as a big-endian
  126. unsigned integer. Next, the server computes 40 bytes of key data
  127. as K = SHA1(g^xy | [00]) | SHA1(g^xy | [01]) where "00" is a single
  128. octet whose value is zero, and "01" is a single octet whose value
  129. is one. The first 16 bytes of K form Kf, and the next 16 bytes of
  130. K form Kb.
  131. Kf is used to encrypt the stream of data going from the OP to the
  132. OR, whereas Kb is used to encrypt the stream of data going from the
  133. OR to the OP.
  134. 4.3. Creating circuits
  135. When creating a circuit through the network, the circuit creator
  136. performs the following steps:
  137. 1. Choose a chain of N onion routers (R_1...R_N) to constitute
  138. the path, such that no router appears in the path twice.
  139. 2. If not already connected to the first router in the chain,
  140. open a new connection to that router.
  141. 3. Choose an ACI not already in use on the connection with the
  142. first router in the chain. If we are an onion router and our
  143. nickname is lexicographically greater than the nickname of the
  144. other side, then let the high bit of the ACI be 1, else 0.
  145. 4. Send a CREATE cell along the connection, to be received by
  146. the first onion router.
  147. 5. Wait until a CREATED cell is received; finish the handshake
  148. and extract the forward key Kf_1 and the back key Kb_1.
  149. 6. For each subsequent onion router R (R_2 through R_N), extend
  150. the circuit to R.
  151. To extend the circuit by a single onion router R_M, the circuit
  152. creator performs these steps:
  153. 1. Create an onion skin, encrypting the RSA-encrypted part with
  154. R's public key.
  155. 2. Encrypt and send the onion skin in a relay EXTEND cell along
  156. the circuit (see section 5).
  157. 3. When a relay EXTENDED cell is received, calculate the shared
  158. keys. The circuit is now extended.
  159. When an onion router receives an EXTEND relay cell, it sends a
  160. CREATE cell to the next onion router, with the enclosed onion skin
  161. as its payload. The initiating onion router chooses some ACI not
  162. yet used on the connection between the two onion routers. (But see
  163. section 4.3. above, concerning choosing ACIs.)
  164. As an extension (called router twins), if the desired next onion
  165. router R in the circuit is down, and some other onion router R'
  166. has the same key as R, then it's ok to extend to R' rather than R.
  167. When an onion router receives a CREATE cell, if it already has a
  168. circuit on the given connection with the given ACI, it drops the
  169. cell. Otherwise, sometime after receiving the CREATE cell, it completes
  170. the DH handshake, and replies with a CREATED cell, containing g^y
  171. as its [128 byte] payload. Upon receiving a CREATED cell, an onion
  172. router packs it payload into an EXTENDED relay cell (see section 5),
  173. and sends that cell up the circuit. Upon receiving the EXTENDED
  174. relay cell, the OP can retrieve g^y.
  175. (As an optimization, OR implementations may delay processing onions
  176. until a break in traffic allows time to do so without harming
  177. network latency too greatly.)
  178. 4.4. Tearing down circuits
  179. Circuits are torn down when an unrecoverable error occurs along
  180. the circuit, or when all streams on a circuit are closed and the
  181. circuit's intended lifetime is over. Circuits may be torn down
  182. either completely or hop-by-hop.
  183. To tear down a circuit completely, an OR or OP sends a DESTROY
  184. cell to the adjacent nodes on that circuit, using the appropriate
  185. direction's ACI.
  186. Upon receiving an outgoing DESTROY cell, an OR frees resources
  187. associated with the corresponding circuit. If it's not the end of
  188. the circuit, it sends a DESTROY cell for that circuit to the next OR
  189. in the circuit. If the node is the end of the circuit, then it tears
  190. down any associated edge connections (see section 5.1).
  191. After a DESTROY cell has been processed, an OR ignores all data or
  192. destroy cells for the corresponding circuit.
  193. To tear down part of a circuit, the OP sends a RELAY_TRUNCATE cell
  194. signaling a given OR (Stream ID zero). That OR sends a DESTROY
  195. cell to the next node in the circuit, and replies to the OP with a
  196. RELAY_TRUNCATED cell.
  197. When an unrecoverable error occurs along one connection in a
  198. circuit, the nodes on either side of the connection should, if they
  199. are able, act as follows: the node closer to the OP should send a
  200. RELAY_TRUNCATED cell towards the OP; the node farther from the OP
  201. should send a DESTROY cell down the circuit.
  202. [We'll have to reevaluate this section once we figure out cleaner
  203. circuit/connection killing conventions. -RD]
  204. 4.5. Routing data cells
  205. When an OR receives a RELAY cell, it checks the cell's ACI and
  206. determines whether it has a corresponding circuit along that
  207. connection. If not, the OR drops the RELAY cell.
  208. Otherwise, if the OR is not at the OP edge of the circuit (that is,
  209. either an 'exit node' or a non-edge node), it de/encrypts the length
  210. field and the payload with AES/CTR, as follows:
  211. 'Forward' relay cell (same direction as CREATE):
  212. Use Kf as key; encrypt.
  213. 'Back' relay cell (opposite direction from CREATE):
  214. Use Kb as key; decrypt.
  215. If the OR recognizes the stream ID on the cell (it is either the ID
  216. of an open stream or the signaling (zero) ID), the OR processes the
  217. contents of the relay cell. Otherwise, it passes the decrypted
  218. relay cell along the circuit if the circuit continues, or drops the
  219. cell if it's the end of the circuit. [Getting an unrecognized
  220. relay cell at the end of the circuit must be allowed for now;
  221. we can reexamine this once we've designed full tcp-style close
  222. handshakes. -RD]
  223. Otherwise, if the data cell is coming from the OP edge of the
  224. circuit, the OP decrypts the length and payload fields with AES/CTR as
  225. follows:
  226. OP sends data cell to node R_M:
  227. For I=1...M, decrypt with Kf_I.
  228. Otherwise, if the data cell is arriving at the OP edge if the
  229. circuit, the OP encrypts the length and payload fields with AES/CTR as
  230. follows:
  231. OP receives data cell:
  232. For I=N...1,
  233. Encrypt with Kb_I. If the stream ID is a recognized
  234. stream for R_I, or if the stream ID is the signaling
  235. ID (zero), then stop and process the payload.
  236. For more information, see section 5 below.
  237. 5. Application connections and stream management
  238. 5.1. Streams
  239. Within a circuit, the OP and the exit node use the contents of
  240. RELAY packets to tunnel end-to-end commands and TCP connections
  241. ("Streams") across circuits. End-to-end commands can be initiated
  242. by either edge; streams are initiated by the OP.
  243. The first 8 bytes of each relay cell are reserved as follows:
  244. Relay command [1 byte]
  245. Stream ID [7 bytes]
  246. The recognized relay commands are:
  247. 1 -- RELAY_BEGIN
  248. 2 -- RELAY_DATA
  249. 3 -- RELAY_END
  250. 4 -- RELAY_CONNECTED
  251. 5 -- RELAY_SENDME
  252. 6 -- RELAY_EXTEND
  253. 7 -- RELAY_EXTENDED
  254. 8 -- RELAY_TRUNCATE
  255. 9 -- RELAY_TRUNCATED
  256. 10 -- RELAY_DROP
  257. All RELAY cells pertaining to the same tunneled stream have the
  258. same stream ID. Stream ID's are chosen randomly by the OP. A
  259. stream ID is considered "recognized" on a circuit C by an OP or an
  260. OR if it already has an existing stream established on that
  261. circuit, or if the stream ID is equal to the signaling stream ID,
  262. which is all zero: [00 00 00 00 00 00 00]
  263. To create a new anonymized TCP connection, the OP sends a
  264. RELAY_BEGIN data cell with a payload encoding the address and port
  265. of the destination host. The stream ID is zero. The payload format is:
  266. NEWSTREAMID | ADDRESS | ':' | PORT | '\000'
  267. where NEWSTREAMID is the newly generated Stream ID to use for
  268. this stream, ADDRESS may be a DNS hostname, or an IPv4 address in
  269. dotted-quad format; and where PORT is encoded in decimal.
  270. Upon receiving this packet, the exit node resolves the address as
  271. necessary, and opens a new TCP connection to the target port. If
  272. the address cannot be resolved, or a connection can't be
  273. established, the exit node replies with a RELAY_END cell.
  274. Otherwise, the exit node replies with a RELAY_CONNECTED cell.
  275. The OP waits for a RELAY_CONNECTED cell before sending any data.
  276. Once a connection has been established, the OP and exit node
  277. package stream data in RELAY_DATA cells, and upon receiving such
  278. cells, echo their contents to the corresponding TCP stream.
  279. Relay RELAY_DROP cells are long-range dummies; upon receiving such
  280. a cell, the OR or OP must drop it.
  281. 5.2. Closing streams
  282. [Note -- TCP streams can only be half-closed for reading. Our
  283. Bickford's conversation was incorrect. -NM]
  284. Because TCP connections can be half-open, we follow an equivalent
  285. to TCP's FIN/FIN-ACK/ACK protocol to close streams.
  286. An exit connection can have a TCP stream in one of three states:
  287. 'OPEN', 'DONE_PACKAGING', and 'DONE_DELIVERING'. For the purposes
  288. of modeling transitions, we treat 'CLOSED' as a fourth state,
  289. although connections in this state are not, in fact, tracked by the
  290. onion router.
  291. A stream begins in the 'OPEN' state. Upon receiving a 'FIN' from
  292. the corresponding TCP connection, the edge node sends a 'RELAY_END'
  293. cell along the circuit and changes its state to 'DONE_PACKAGING'.
  294. Upon receiving a 'RELAY_END' cell, an edge node sends a 'FIN' to
  295. the corresponding TCP connection (e.g., by calling
  296. shutdown(SHUT_WR)) and changing its state to 'DONE_DELIVERING'.
  297. When a stream in already in 'DONE_DELIVERING' receives a 'FIN', it
  298. also sends a 'RELAY_END' along the circuit, and changes its state
  299. to 'CLOSED'. When a stream already in 'DONE_PACKAGING' receives a
  300. 'RELAY_END' cell, it sends a 'FIN' and changes its state to
  301. 'CLOSED'.
  302. [Note: Please rename 'RELAY_END2'. :) -NM ]
  303. If an edge node encounters an error on any stram, it sends a
  304. 'RELAY_END2' cell along the circuit (if possible) and closes the
  305. TCP connection immediately. If an edge node receives a
  306. 'RELAY_END2' cell for any stream, it closes the TCP connection
  307. completely, and sends nothing along the circuit.
  308. 6. Flow control
  309. 6.1. Link throttling
  310. Each node should do appropriate bandwidth throttling to keep its
  311. user happy.
  312. Communicants rely on TCP's default flow control to push back when they
  313. stop reading.
  314. 6.2. Link padding
  315. Currently nodes are not required to do any sort of link padding or
  316. dummy traffic. Because strong attacks exist even with link padding,
  317. and because link padding greatly increases the bandwidth requirements
  318. for running a node, we plan to leave out link padding until this
  319. tradeoff is better understood.
  320. 6.3. Circuit-level flow control
  321. To control a circuit's bandwidth usage, each OR keeps track of
  322. two 'windows', consisting of how many RELAY_DATA cells it is
  323. allowed to package for transmission, and how many RELAY_DATA cells
  324. it is willing to deliver to streams outside the network.
  325. Each 'window' value is initially set to 1000 data cells
  326. in each direction (cells that are not data cells do not affect
  327. the window). When an OR is willing to deliver more cells, it sends a
  328. RELAY_SENDME cell towards the OP, with Stream ID zero. When an OR
  329. receives a RELAY_SENDME cell with stream ID zero, it increments its
  330. packaging window.
  331. Either of these cells increment the corresponding window by 100.
  332. The OP behaves identically, except that it must track a packaging
  333. window and a delivery window for every OR in the circuit.
  334. An OR or OP sends cells to increment its delivery window when the
  335. corresponding window value falls under some threshold (900).
  336. If a packaging window reaches 0, the OR or OP stops reading from
  337. TCP connections for all streams on the corresponding circuit, and
  338. sends no more RELAY_DATA cells until receiving a RELAY_SENDME cell.
  339. 6.4. Stream-level flow control
  340. Edge nodes use RELAY_SENDME cells to implement end-to-end flow
  341. control for individual connections across circuits. Similarly to
  342. circuit-level flow control, edge nodes begin with a window of cells
  343. (500) per stream, and increment the window by a fixed value (50)
  344. upon receiving a RELAY_SENDME cell. Edge nodes initiate RELAY_SENDME
  345. cells when both a) the window is <= 450, and b) there are less than
  346. ten cell payloads remaining to be flushed at that edge.
  347. 7. Directories and routers
  348. 7.1. Router descriptor format.
  349. (Unless otherwise noted, tokens on the same line are space-separated.)
  350. Router ::= Router-Line Date-Line Onion-Key Link-Key Signing-Key Exit-Policy Router-Signature NL
  351. Router-Line ::= "router" nickname address ORPort SocksPort DirPort bandwidth NL
  352. Date-Line ::= "published" YYYY-MM-DD HH:MM:SS NL
  353. Onion-key ::= "onion-key" NL a public key in PEM format NL
  354. Link-key ::= "link-key" NL a public key in PEM format NL
  355. Signing-Key ::= "signing-key" NL a public key in PEM format NL
  356. Exit-Policy ::= Exit-Line*
  357. Exit-Line ::= ("accept"|"reject") string NL
  358. Router-Signature ::= "router-signature" NL Signature
  359. Signature ::= "-----BEGIN SIGNATURE-----" NL
  360. Base-64-encoded-signature NL "-----END SIGNATURE-----" NL
  361. ORport ::= port where the router listens for routers/proxies (speaking cells)
  362. SocksPort ::= where the router listens for applications (speaking socks)
  363. DirPort ::= where the router listens for directory download requests
  364. bandwidth ::= maximum bandwidth, in bytes/s
  365. nickname ::= between 1 and 32 alphanumeric characters. case-insensitive.
  366. Example:
  367. router moria1 moria.mit.edu 9001 9021 9031 100000
  368. published 2003-09-24 19:36:05
  369. -----BEGIN RSA PUBLIC KEY-----
  370. MIGJAoGBAMBBuk1sYxEg5jLAJy86U3GGJ7EGMSV7yoA6mmcsEVU3pwTUrpbpCmwS
  371. 7BvovoY3z4zk63NZVBErgKQUDkn3pp8n83xZgEf4GI27gdWIIwaBjEimuJlEY+7K
  372. nZ7kVMRoiXCbjL6VAtNa4Zy1Af/GOm0iCIDpholeujQ95xew7rQnAgMA//8=
  373. -----END RSA PUBLIC KEY-----
  374. signing-key
  375. -----BEGIN RSA PUBLIC KEY-----
  376. 7BvovoY3z4zk63NZVBErgKQUDkn3pp8n83xZgEf4GI27gdWIIwaBjEimuJlEY+7K
  377. MIGJAoGBAMBBuk1sYxEg5jLAJy86U3GGJ7EGMSV7yoA6mmcsEVU3pwTUrpbpCmwS
  378. f/GOm0iCIDpholeujQ95xew7rnZ7kVMRoiXCbjL6VAtNa4Zy1AQnAgMA//8=
  379. -----END RSA PUBLIC KEY-----
  380. reject 18.0.0.0/24
  381. Note: The extra newline at the end of the router block is intentional.
  382. 7.2. Directory format
  383. Directory ::= Directory-Header Directory-Router Router* Signature
  384. Directory-Header ::= "signed-directory" NL Software-Line NL
  385. Software-Line: "recommended-software" comma-separated-version-list
  386. Directory-Router ::= Router
  387. Directory-Signature ::= "directory-signature" NL Signature
  388. Signature ::= "-----BEGIN SIGNATURE-----" NL
  389. Base-64-encoded-signature NL "-----END SIGNATURE-----" NL
  390. Note: The router block for the directory server must appear first.
  391. The signature is computed by computing the SHA-1 hash of the
  392. directory, from the characters "signed-directory", through the newline
  393. after "directory-signature". This digest is then padded with PKCS.1,
  394. and signed with the directory server's signing key.
  395. 7.3. Behavior of a directory server
  396. lists nodes that are connected currently
  397. speaks http on a socket, spits out directory on request
  398. -----------
  399. (for emacs)
  400. Local Variables:
  401. mode:text
  402. indent-tabs-mode:nil
  403. fill-column:77
  404. End: