tor-spec.txt 41 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612613614615616617618619620621622623624625626627628629630631632633634635636637638639640641642643644645646647648649650651652653654655656657658659660661662663664665666667668669670671672673674675676677678679680681682683684685686687688689690691692693694695696697698699700701702703704705706707708709710711712713714715716717718719720721722723724725726727728729730731732733734735736737738739740741742743744745746747748749750751752753754755756757758759760761762763764765766767768769770771772773774775776777778779780781782783784785786787788789790791792793794795796797798799800801802803804805806807808809810811812813814815816817818819820821822823824825826827828829830831832833834835836837838839840841842843844845846847848849850851852853854855856857858859860861862863864865866867868869870871872873874875876877878879880881882883884885886887888889890891892893894895896897898899900901902903904905906907908909910911912913914915916917918919920921922923924925926927928929930931932933934935
  1. $Id$
  2. Tor Protocol Specification
  3. Roger Dingledine
  4. Nick Mathewson
  5. Note: This document aims to specify Tor as implemented in 0.1.2.1-alpha-dev
  6. and later. Future versions of Tor will implement improved protocols, and
  7. compatibility is not guaranteed.
  8. THIS DOCUMENT IS UNSTABLE. Right now, we're revising the protocol to remove
  9. a few long-standing limitations. For the most stable current version of the
  10. protocol, see tor-spec-v0.txt; current versions of Tor are backward-compatible.
  11. This specification is not a design document; most design criteria
  12. are not examined. For more information on why Tor acts as it does,
  13. see tor-design.pdf.
  14. TODO for v1 revision:
  15. - Fix onionskin handshake scheme to be more mainstream, less nutty.
  16. Can we just do
  17. E(HMAC(g^x), g^x) rather than just E(g^x) ?
  18. No, that has the same flaws as before. We should send
  19. E(g^x, C) with random C and expect g^y, HMAC_C(K=g^xy).
  20. Better ask Ian; probably Stephen too.
  21. - Versioned CREATE and friends
  22. - Length on CREATE and friends
  23. - Versioning on circuits
  24. TODO:
  25. - REASON_CONNECTFAILED should include an IP.
  26. - Copy prose from tor-design to make everything more readable.
  27. - Spec when we should rotate which keys (tls, link, etc)?
  28. 0. Preliminaries
  29. 0.1. Notation and encoding
  30. PK -- a public key.
  31. SK -- a private key.
  32. K -- a key for a symmetric cypher.
  33. a|b -- concatenation of 'a' and 'b'.
  34. [A0 B1 C2] -- a three-byte sequence, containing the bytes with
  35. hexadecimal values A0, B1, and C2, in that order.
  36. All numeric values are encoded in network (big-endian) order.
  37. H(m) -- a cryptographic hash of m.
  38. 0.2. Security parameters
  39. Tor uses a stream cipher, a public-key cipher, the Diffie-Hellman
  40. protocol, and a hash function.
  41. KEY_LEN -- the length of the stream cipher's key, in bytes.
  42. PK_ENC_LEN -- the length of a public-key encrypted message, in bytes.
  43. PK_PAD_LEN -- the number of bytes added in padding for public-key
  44. encryption, in bytes. (The largest number of bytes that can be encrypted
  45. in a single public-key operation is therefore PK_ENC_LEN-PK_PAD_LEN.)
  46. DH_LEN -- the number of bytes used to represent a member of the
  47. Diffie-Hellman group.
  48. DH_SEC_LEN -- the number of bytes used in a Diffie-Hellman private key (x).
  49. HASH_LEN -- the length of the hash function's output, in bytes.
  50. PAYLOAD_LEN -- The longest allowable cell payload, in bytes. (509)
  51. CELL_LEN -- The length of a Tor cell, in bytes.
  52. 0.3. Ciphers
  53. For a stream cipher, we use 128-bit AES in counter mode, with an IV of all
  54. 0 bytes.
  55. For a public-key cipher, we use RSA with 1024-bit keys and a fixed
  56. exponent of 65537. We use OAEP padding, with SHA-1 as its digest
  57. function. (For OAEP padding, see
  58. ftp://ftp.rsasecurity.com/pub/pkcs/pkcs-1/pkcs-1v2-1.pdf)
  59. For Diffie-Hellman, we use a generator (g) of 2. For the modulus (p), we
  60. use the 1024-bit safe prime from rfc2409 section 6.2 whose hex
  61. representation is:
  62. "FFFFFFFFFFFFFFFFC90FDAA22168C234C4C6628B80DC1CD129024E08"
  63. "8A67CC74020BBEA63B139B22514A08798E3404DDEF9519B3CD3A431B"
  64. "302B0A6DF25F14374FE1356D6D51C245E485B576625E7EC6F44C42E9"
  65. "A637ED6B0BFF5CB6F406B7EDEE386BFB5A899FA5AE9F24117C4B1FE6"
  66. "49286651ECE65381FFFFFFFFFFFFFFFF"
  67. As an optimization, implementations SHOULD choose DH private keys (x) of
  68. 320 bits. Implementations that do this MUST never use any DH key more
  69. than once.
  70. [May other implementations reuse their DH keys?? -RD]
  71. For a hash function, we use SHA-1.
  72. KEY_LEN=16.
  73. DH_LEN=128; DH_GROUP_LEN=40.
  74. PK_ENC_LEN=128; PK_PAD_LEN=42.
  75. HASH_LEN=20.
  76. When we refer to "the hash of a public key", we mean the SHA-1 hash of the
  77. DER encoding of an ASN.1 RSA public key (as specified in PKCS.1).
  78. All "random" values should be generated with a cryptographically strong
  79. random number generator, unless otherwise noted.
  80. The "hybrid encryption" of a byte sequence M with a public key PK is
  81. computed as follows:
  82. 1. If M is less than PK_ENC_LEN-PK_PAD_LEN, pad and encrypt M with PK.
  83. 2. Otherwise, generate a KEY_LEN byte random key K.
  84. Let M1 = the first PK_ENC_LEN-PK_PAD_LEN-KEY_LEN bytes of M,
  85. and let M2 = the rest of M.
  86. Pad and encrypt K|M1 with PK. Encrypt M2 with our stream cipher,
  87. using the key K. Concatenate these encrypted values.
  88. [XXX Note that this "hybrid encryption" approach does not prevent
  89. an attacker from adding or removing bytes to the end of M. It also
  90. allows attackers to modify the bytes not covered by the OAEP --
  91. see Goldberg's PET2006 paper for details. We will add a MAC to this
  92. scheme one day. -RD]
  93. 0.4. Other parameter values
  94. CELL_LEN=512
  95. 1. System overview
  96. Tor is a distributed overlay network designed to anonymize
  97. low-latency TCP-based applications such as web browsing, secure shell,
  98. and instant messaging. Clients choose a path through the network and
  99. build a ``circuit'', in which each node (or ``onion router'' or ``OR'')
  100. in the path knows its predecessor and successor, but no other nodes in
  101. the circuit. Traffic flowing down the circuit is sent in fixed-size
  102. ``cells'', which are unwrapped by a symmetric key at each node (like
  103. the layers of an onion) and relayed downstream.
  104. 1.1. Protocol Versioning
  105. The node-to-node TLS-based "OR connection" protocol and the multi-hop
  106. "circuit" protocol are versioned quasi-independently. (Certain versions
  107. of the circuit protocol may require a minimum version of the connection
  108. protocol to be used.)
  109. Version numbers are incremented for backward-incompatible protocol changes
  110. only. Backward-compatible changes are generally implemented by adding
  111. additional fields to existing structures; implementations MUST ignore
  112. fields they do not expect.
  113. Parties negotiate OR connection versions as described below in sections
  114. 4.1 and 4.2.
  115. 2. Connections
  116. Tor uses TLS for link encryption. All implementations MUST support
  117. the TLS ciphersuite "TLS_EDH_RSA_WITH_DES_192_CBC3_SHA", and SHOULD
  118. support "TLS_DHE_RSA_WITH_AES_128_CBC_SHA" if it is available.
  119. Implementations MAY support other ciphersuites, but MUST NOT
  120. support any suite without ephemeral keys, symmetric keys of at
  121. least KEY_LEN bits, and digests of at least HASH_LEN bits.
  122. Even though the connection protocol is identical, we think of the
  123. initiator as either an onion router (OR) if it is willing to relay
  124. traffic for other Tor users, or an onion proxy (OP) if it only handles
  125. local requests. Onion proxies SHOULD NOT provide long-term-trackable
  126. identifiers in their handshakes.
  127. The connection initiator always sends a two-certificate chain,
  128. consisting of a
  129. certificate using a short-term connection key and a second, self-
  130. signed certificate containing the OR's identity key. The commonName of the
  131. first certificate is the OR's nickname, and the commonName of the second
  132. certificate is the OR's nickname, followed by a space and the string
  133. "<identity>".
  134. Implementations running 0.2.1.0-alpha-dev and earlier used an
  135. organizationName of "Tor" or "TOR". Current implementations (which
  136. support the version negotiation protocol in section 4.1) MUST NOT
  137. have either of these values for their organizationName.
  138. All parties receiving certificates must confirm that the identity key is
  139. as expected. (When initiating a connection, the expected identity key is
  140. the one given in the directory; when creating a connection because of an
  141. EXTEND cell, the expected identity key is the one given in the cell.) If
  142. the key is not as expected, the party must close the connection.
  143. All parties SHOULD reject connections to or from ORs that have malformed
  144. or missing certificates. ORs MAY accept or reject connections from OPs
  145. with malformed or missing certificates.
  146. Once a TLS connection is established, the two sides send cells
  147. (specified below) to one another. Cells are sent serially. All
  148. cells are CELL_LEN bytes long. Cells may be sent embedded in TLS
  149. records of any size or divided across TLS records, but the framing
  150. of TLS records MUST NOT leak information about the type or contents
  151. of the cells.
  152. TLS connections are not permanent. Either side may close a connection
  153. if there are no circuits running over it and an amount of time
  154. (KeepalivePeriod, defaults to 5 minutes) has passed.
  155. (As an exception, directory servers may try to stay connected to all of
  156. the ORs -- though this will be phased out for the Tor 0.1.2.x release.)
  157. 3. Cell Packet format
  158. The basic unit of communication for onion routers and onion
  159. proxies is a fixed-width "cell".
  160. On a version 0 connection, each cell contains the following
  161. fields:
  162. CircID [2 bytes]
  163. Command [1 byte]
  164. Payload (padded with 0 bytes) [PAYLOAD_LEN bytes]
  165. On a version 1 connection, each cell contains the following fields:
  166. CircID [3 bytes]
  167. Command [1 byte]
  168. Payload (padded with 0 bytes) [PAYLOAD_LEN bytes]
  169. The CircID field determines which circuit, if any, the cell is
  170. associated with.
  171. The 'Command' field holds one of the following values:
  172. 0 -- PADDING (Padding) (See Sec 7.2)
  173. 1 -- CREATE (Create a circuit) (See Sec 5.1)
  174. 2 -- CREATED (Acknowledge create) (See Sec 5.1)
  175. 3 -- RELAY (End-to-end data) (See Sec 5.5 and 6)
  176. 4 -- DESTROY (Stop using a circuit) (See Sec 5.4)
  177. 5 -- CREATE_FAST (Create a circuit, no PK) (See Sec 5.1)
  178. 6 -- CREATED_FAST (Circuit created, no PK) (See Sec 5.1)
  179. 7 -- VERSIONS (Negotiate versions) (See Sec 4.1)
  180. 8 -- NETINFO (Time and MITM-prevention) (See Sec 4.2)
  181. The interpretation of 'Payload' depends on the type of the cell.
  182. PADDING: Payload is unused.
  183. CREATE: Payload contains the handshake challenge.
  184. CREATED: Payload contains the handshake response.
  185. RELAY: Payload contains the relay header and relay body.
  186. DESTROY: Payload contains a reason for closing the circuit.
  187. (see 5.4)
  188. Upon receiving any other value for the command field, an OR must
  189. drop the cell. [XXXX Versions prior to 0.1.0.?? logged a warning
  190. when dropping the cell; this is bad behavior. -NM]
  191. The payload is padded with 0 bytes.
  192. PADDING cells are currently used to implement connection keepalive.
  193. If there is no other traffic, ORs and OPs send one another a PADDING
  194. cell every few minutes.
  195. CREATE, CREATED, and DESTROY cells are used to manage circuits;
  196. see section 4 below.
  197. RELAY cells are used to send commands and data along a circuit; see
  198. section 5 below.
  199. VERSIONS cells are used to introduce parameters and characteristics of
  200. Tor clients and servers when connections are established.
  201. 4, Connection management
  202. Upon establishing a TLS connection, both parties immediately begin
  203. negotiating a connection protocol version and other connection parameters.
  204. 4.1. VERSIONS cells
  205. When a Tor connection is established, both parties normally send a
  206. VERSIONS cell before sending any other cells. (But see below.)
  207. NumVersions [1 byte]
  208. Versions [NumVersions bytes]
  209. "Versions" is a sequence of NumVersions link connection protocol versions,
  210. each one byte long. Parties should list all of the versions which they
  211. are able and willing to support. Parties can only communicate if they
  212. have some connection protocol version in common.
  213. Version 0.1.x.y-alpha and earlier don't understand VERSIONS cells,
  214. and therefore don't support version negotiation. Thus, waiting until
  215. the other side has sent a VERSIONS cell won't work for these servers:
  216. if they send no cells back, it is impossible to tell whether they
  217. have sent a VERSIONS cell that has been stalled, or whether they have
  218. dropped our own VERSIONS cell as unrecognized. Thus, immediately after
  219. a TLS connection has been established, the parties check whether the
  220. other side has an obsolete certificate (organizationName equal to "Tor"
  221. or "TOR"). If the other party presented an obsolete certificate,
  222. we assume a v0 connection. Otherwise, both parties send VERSIONS
  223. cells listing all their supported versions. Upon receiving the
  224. other party's VERSIONS cell, the implementation begins using the
  225. highest-valued version common to both cells. If the first cell from
  226. the other party is _not_ a VERSIONS cell, we assume a v0 protocol.
  227. Implementations MUST discard cells that are not the first cells sent on a
  228. connection.
  229. 4.2. MITM-prevention and time checking
  230. If we negotiate a v1 connection or higher, the first cell we send SHOULD
  231. be a NETINFO cell. Implementations SHOULD NOT send NETINFO cells at other
  232. times.
  233. A NETINFO cell contains:
  234. Timestamp [4 bytes]
  235. This OR's address [variable]
  236. Other OR's address [variable]
  237. Timestamp is the OR's current Unix time, in seconds since the epoch. If
  238. an implementation receives time values from many validated ORs that
  239. indicate that its clock is skewed, it SHOULD try to warn the
  240. administrator.
  241. Each address contains Type/Length/Value as used in Section 6.4. The first
  242. address is the address of the interface the party sending the VERSIONS cell
  243. used to connect to or accept connections from the other -- we include it
  244. to block a man-in-the-middle attack on TLS that lets an attacker bounce
  245. traffic through his own computers to enable timing and packet-counting
  246. attacks.
  247. The second address is the one that the party sending the VERSIONS cell
  248. believes the other has -- it can be used to learn what your IP address
  249. is if you have no other hints.
  250. 5. Circuit management
  251. 5.1. CREATE and CREATED cells
  252. Users set up circuits incrementally, one hop at a time. To create a
  253. new circuit, OPs send a CREATE cell to the first node, with the
  254. first half of the DH handshake; that node responds with a CREATED
  255. cell with the second half of the DH handshake plus the first 20 bytes
  256. of derivative key data (see section 5.2). To extend a circuit past
  257. the first hop, the OP sends an EXTEND relay cell (see section 5)
  258. which instructs the last node in the circuit to send a CREATE cell
  259. to extend the circuit.
  260. The payload for a CREATE cell is an 'onion skin', which consists
  261. of the first step of the DH handshake data (also known as g^x).
  262. This value is hybrid-encrypted (see 0.3) to Bob's public key, giving
  263. an onion-skin of:
  264. PK-encrypted:
  265. Padding padding [PK_PAD_LEN bytes]
  266. Symmetric key [KEY_LEN bytes]
  267. First part of g^x [PK_ENC_LEN-PK_PAD_LEN-KEY_LEN bytes]
  268. Symmetrically encrypted:
  269. Second part of g^x [DH_LEN-(PK_ENC_LEN-PK_PAD_LEN-KEY_LEN)
  270. bytes]
  271. The relay payload for an EXTEND relay cell consists of:
  272. Address [4 bytes]
  273. Port [2 bytes]
  274. Onion skin [DH_LEN+KEY_LEN+PK_PAD_LEN bytes]
  275. Identity fingerprint [HASH_LEN bytes]
  276. The port and address field denote the IPV4 address and port of the next
  277. onion router in the circuit; the public key hash is the hash of the PKCS#1
  278. ASN1 encoding of the next onion router's identity (signing) key. (See 0.3
  279. above.) (Including this hash allows the extending OR verify that it is
  280. indeed connected to the correct target OR, and prevents certain
  281. man-in-the-middle attacks.)
  282. The payload for a CREATED cell, or the relay payload for an
  283. EXTENDED cell, contains:
  284. DH data (g^y) [DH_LEN bytes]
  285. Derivative key data (KH) [HASH_LEN bytes] <see 5.2 below>
  286. The CircID for a CREATE cell is an arbitrarily chosen 2-byte integer,
  287. selected by the node (OP or OR) that sends the CREATE cell. To prevent
  288. CircID collisions, when one OR sends a CREATE cell to another, it chooses
  289. from only one half of the possible values based on the ORs' public
  290. identity keys: if the sending OR has a lower key, it chooses a CircID with
  291. an MSB of 0; otherwise, it chooses a CircID with an MSB of 1.
  292. Public keys are compared numerically by modulus.
  293. As usual with DH, x and y MUST be generated randomly.
  294. [
  295. To implement backward-compatible version negotiation, parties MUST
  296. drop CREATE cells with all-[00] onion-skins.
  297. ]
  298. 5.1.1. CREATE_FAST/CREATED_FAST cells
  299. When initializing the first hop of a circuit, the OP has already
  300. established the OR's identity and negotiated a secret key using TLS.
  301. Because of this, it is not always necessary for the OP to perform the
  302. public key operations to create a circuit. In this case, the
  303. OP MAY send a CREATE_FAST cell instead of a CREATE cell for the first
  304. hop only. The OR responds with a CREATED_FAST cell, and the circuit is
  305. created.
  306. A CREATE_FAST cell contains:
  307. Key material (X) [HASH_LEN bytes]
  308. A CREATED_FAST cell contains:
  309. Key material (Y) [HASH_LEN bytes]
  310. Derivative key data [HASH_LEN bytes] (See 5.2 below)
  311. The values of X and Y must be generated randomly.
  312. [Versions of Tor before 0.1.0.6-rc did not support these cell types;
  313. clients should not send CREATE_FAST cells to older Tor servers.]
  314. If an OR sees a circuit created with CREATE_FAST, the OR is sure to be the
  315. first hop of a circuit. ORs SHOULD reject attempts to create streams with
  316. RELAY_BEGIN exiting the circuit at the first hop: letting Tor be used as a
  317. single hop proxy makes exit nodes a more attractive target for compromise.
  318. 5.2. Setting circuit keys
  319. Once the handshake between the OP and an OR is completed, both can
  320. now calculate g^xy with ordinary DH. Before computing g^xy, both client
  321. and server MUST verify that the received g^x or g^y value is not degenerate;
  322. that is, it must be strictly greater than 1 and strictly less than p-1
  323. where p is the DH modulus. Implementations MUST NOT complete a handshake
  324. with degenerate keys. Implementations MUST NOT discard other "weak"
  325. g^x values.
  326. (Discarding degenerate keys is critical for security; if bad keys
  327. are not discarded, an attacker can substitute the server's CREATED
  328. cell's g^y with 0 or 1, thus creating a known g^xy and impersonating
  329. the server. Discarding other keys may allow attacks to learn bits of
  330. the private key.)
  331. (The mainline Tor implementation, in the 0.1.1.x-alpha series, discarded
  332. all g^x values less than 2^24, greater than p-2^24, or having more than
  333. 1024-16 identical bits. This served no useful purpose, and we stopped.)
  334. If CREATE or EXTEND is used to extend a circuit, the client and server
  335. base their key material on K0=g^xy, represented as a big-endian unsigned
  336. integer.
  337. If CREATE_FAST is used, the client and server base their key material on
  338. K0=X|Y.
  339. From the base key material K0, they compute KEY_LEN*2+HASH_LEN*3 bytes of
  340. derivative key data as
  341. K = H(K0 | [00]) | H(K0 | [01]) | H(K0 | [02]) | ...
  342. The first HASH_LEN bytes of K form KH; the next HASH_LEN form the forward
  343. digest Df; the next HASH_LEN 41-60 form the backward digest Db; the next
  344. KEY_LEN 61-76 form Kf, and the final KEY_LEN form Kb. Excess bytes from K
  345. are discarded.
  346. KH is used in the handshake response to demonstrate knowledge of the
  347. computed shared key. Df is used to seed the integrity-checking hash
  348. for the stream of data going from the OP to the OR, and Db seeds the
  349. integrity-checking hash for the data stream from the OR to the OP. Kf
  350. is used to encrypt the stream of data going from the OP to the OR, and
  351. Kb is used to encrypt the stream of data going from the OR to the OP.
  352. 5.3. Creating circuits
  353. When creating a circuit through the network, the circuit creator
  354. (OP) performs the following steps:
  355. 1. Choose an onion router as an exit node (R_N), such that the onion
  356. router's exit policy includes at least one pending stream that
  357. needs a circuit (if there are any).
  358. 2. Choose a chain of (N-1) onion routers
  359. (R_1...R_N-1) to constitute the path, such that no router
  360. appears in the path twice.
  361. 3. If not already connected to the first router in the chain,
  362. open a new connection to that router.
  363. 4. Choose a circID not already in use on the connection with the
  364. first router in the chain; send a CREATE cell along the
  365. connection, to be received by the first onion router.
  366. 5. Wait until a CREATED cell is received; finish the handshake
  367. and extract the forward key Kf_1 and the backward key Kb_1.
  368. 6. For each subsequent onion router R (R_2 through R_N), extend
  369. the circuit to R.
  370. To extend the circuit by a single onion router R_M, the OP performs
  371. these steps:
  372. 1. Create an onion skin, encrypted to R_M's public key.
  373. 2. Send the onion skin in a relay EXTEND cell along
  374. the circuit (see section 5).
  375. 3. When a relay EXTENDED cell is received, verify KH, and
  376. calculate the shared keys. The circuit is now extended.
  377. When an onion router receives an EXTEND relay cell, it sends a CREATE
  378. cell to the next onion router, with the enclosed onion skin as its
  379. payload. The initiating onion router chooses some circID not yet
  380. used on the connection between the two onion routers. (But see
  381. section 5.1. above, concerning choosing circIDs based on
  382. lexicographic order of nicknames.)
  383. When an onion router receives a CREATE cell, if it already has a
  384. circuit on the given connection with the given circID, it drops the
  385. cell. Otherwise, after receiving the CREATE cell, it completes the
  386. DH handshake, and replies with a CREATED cell. Upon receiving a
  387. CREATED cell, an onion router packs it payload into an EXTENDED relay
  388. cell (see section 5), and sends that cell up the circuit. Upon
  389. receiving the EXTENDED relay cell, the OP can retrieve g^y.
  390. (As an optimization, OR implementations may delay processing onions
  391. until a break in traffic allows time to do so without harming
  392. network latency too greatly.)
  393. 5.4. Tearing down circuits
  394. Circuits are torn down when an unrecoverable error occurs along
  395. the circuit, or when all streams on a circuit are closed and the
  396. circuit's intended lifetime is over. Circuits may be torn down
  397. either completely or hop-by-hop.
  398. To tear down a circuit completely, an OR or OP sends a DESTROY
  399. cell to the adjacent nodes on that circuit, using the appropriate
  400. direction's circID.
  401. Upon receiving an outgoing DESTROY cell, an OR frees resources
  402. associated with the corresponding circuit. If it's not the end of
  403. the circuit, it sends a DESTROY cell for that circuit to the next OR
  404. in the circuit. If the node is the end of the circuit, then it tears
  405. down any associated edge connections (see section 6.1).
  406. After a DESTROY cell has been processed, an OR ignores all data or
  407. destroy cells for the corresponding circuit.
  408. To tear down part of a circuit, the OP may send a RELAY_TRUNCATE cell
  409. signaling a given OR (Stream ID zero). That OR sends a DESTROY
  410. cell to the next node in the circuit, and replies to the OP with a
  411. RELAY_TRUNCATED cell.
  412. When an unrecoverable error occurs along one connection in a
  413. circuit, the nodes on either side of the connection should, if they
  414. are able, act as follows: the node closer to the OP should send a
  415. RELAY_TRUNCATED cell towards the OP; the node farther from the OP
  416. should send a DESTROY cell down the circuit.
  417. The payload of a RELAY_TRUNCATED or DESTROY cell contains a single octet,
  418. describing why the circuit is being closed or truncated. When sending a
  419. TRUNCATED or DESTROY cell because of another TRUNCATED or DESTROY cell,
  420. the error code should be propagated. The origin of a circuit always sets
  421. this error code to 0, to avoid leaking its version.
  422. The error codes are:
  423. 0 -- NONE (No reason given.)
  424. 1 -- PROTOCOL (Tor protocol violation.)
  425. 2 -- INTERNAL (Internal error.)
  426. 3 -- REQUESTED (A client sent a TRUNCATE command.)
  427. 4 -- HIBERNATING (Not currently operating; trying to save bandwidth.)
  428. 5 -- RESOURCELIMIT (Out of memory, sockets, or circuit IDs.)
  429. 6 -- CONNECTFAILED (Unable to reach server.)
  430. 7 -- OR_IDENTITY (Connected to server, but its OR identity was not
  431. as expected.)
  432. 8 -- OR_CONN_CLOSED (The OR connection that was carrying this circuit
  433. died.)
  434. 9 -- FINISHED (The circuit has expired for being dirty or old.)
  435. 10 -- TIMEOUT (Circuit construction took too long)
  436. 11 -- DESTROYED (The circuit was destroyed w/o client TRUNCATE)
  437. 12 -- NOSUCHSERVICE (Request for unknown hidden service)
  438. [Versions of Tor prior to 0.1.0.11 didn't send reasons; implementations
  439. MUST accept empty TRUNCATED and DESTROY cells.]
  440. 5.5. Routing relay cells
  441. When an OR receives a RELAY cell, it checks the cell's circID and
  442. determines whether it has a corresponding circuit along that
  443. connection. If not, the OR drops the RELAY cell.
  444. Otherwise, if the OR is not at the OP edge of the circuit (that is,
  445. either an 'exit node' or a non-edge node), it de/encrypts the payload
  446. with the stream cipher, as follows:
  447. 'Forward' relay cell (same direction as CREATE):
  448. Use Kf as key; decrypt.
  449. 'Back' relay cell (opposite direction from CREATE):
  450. Use Kb as key; encrypt.
  451. Note that in counter mode, decrypt and encrypt are the same operation.
  452. The OR then decides whether it recognizes the relay cell, by
  453. inspecting the payload as described in section 6.1 below. If the OR
  454. recognizes the cell, it processes the contents of the relay cell.
  455. Otherwise, it passes the decrypted relay cell along the circuit if
  456. the circuit continues. If the OR at the end of the circuit
  457. encounters an unrecognized relay cell, an error has occurred: the OR
  458. sends a DESTROY cell to tear down the circuit.
  459. When a relay cell arrives at an OP, the OP decrypts the payload
  460. with the stream cipher as follows:
  461. OP receives data cell:
  462. For I=N...1,
  463. Decrypt with Kb_I. If the payload is recognized (see
  464. section 6..1), then stop and process the payload.
  465. For more information, see section 6 below.
  466. 6. Application connections and stream management
  467. 6.1. Relay cells
  468. Within a circuit, the OP and the exit node use the contents of
  469. RELAY packets to tunnel end-to-end commands and TCP connections
  470. ("Streams") across circuits. End-to-end commands can be initiated
  471. by either edge; streams are initiated by the OP.
  472. The payload of each unencrypted RELAY cell consists of:
  473. Relay command [1 byte]
  474. 'Recognized' [2 bytes]
  475. StreamID [2 bytes]
  476. Digest [4 bytes]
  477. Length [2 bytes]
  478. Data [CELL_LEN-14 bytes]
  479. The relay commands are:
  480. 1 -- RELAY_BEGIN [forward]
  481. 2 -- RELAY_DATA [forward or backward]
  482. 3 -- RELAY_END [forward or backward]
  483. 4 -- RELAY_CONNECTED [backward]
  484. 5 -- RELAY_SENDME [forward or backward] [sometimes control]
  485. 6 -- RELAY_EXTEND [forward] [control]
  486. 7 -- RELAY_EXTENDED [backward] [control]
  487. 8 -- RELAY_TRUNCATE [forward] [control]
  488. 9 -- RELAY_TRUNCATED [backward] [control]
  489. 10 -- RELAY_DROP [forward or backward] [control]
  490. 11 -- RELAY_RESOLVE [forward]
  491. 12 -- RELAY_RESOLVED [backward]
  492. 13 -- RELAY_BEGIN_DIR [forward]
  493. Commands labelled as "forward" must only be sent by the originator
  494. of the circuit. Commands labelled as "backward" must only be sent by
  495. other nodes in the circuit back to the originator. Commands marked
  496. as either can be sent either by the originator or other nodes.
  497. The 'recognized' field in any unencrypted relay payload is always set
  498. to zero; the 'digest' field is computed as the first four bytes of
  499. the running digest of all the bytes that have been destined for
  500. this hop of the circuit or originated from this hop of the circuit,
  501. seeded from Df or Db respectively (obtained in section 5.2 above),
  502. and including this RELAY cell's entire payload (taken with the digest
  503. field set to zero).
  504. When the 'recognized' field of a RELAY cell is zero, and the digest
  505. is correct, the cell is considered "recognized" for the purposes of
  506. decryption (see section 5.5 above).
  507. (The digest does not include any bytes from relay cells that do
  508. not start or end at this hop of the circuit. That is, it does not
  509. include forwarded data. Therefore if 'recognized' is zero but the
  510. digest does not match, the running digest at that node should
  511. not be updated, and the cell should be forwarded on.)
  512. All RELAY cells pertaining to the same tunneled stream have the
  513. same stream ID. StreamIDs are chosen arbitrarily by the OP. RELAY
  514. cells that affect the entire circuit rather than a particular
  515. stream use a StreamID of zero -- they are marked in the table above
  516. as "[control]" style cells. (Sendme cells are marked as "sometimes
  517. control" because they can take include a StreamID or not depending
  518. on their purpose -- see Section 7.)
  519. The 'Length' field of a relay cell contains the number of bytes in
  520. the relay payload which contain real payload data. The remainder of
  521. the payload is padded with NUL bytes.
  522. If the RELAY cell is recognized but the relay command is not
  523. understood, the cell must be dropped and ignored. Its contents
  524. still count with respect to the digests, though. [Before
  525. 0.1.1.10, Tor closed circuits when it received an unknown relay
  526. command. Perhaps this will be more forward-compatible. -RD]
  527. 6.2. Opening streams and transferring data
  528. To open a new anonymized TCP connection, the OP chooses an open
  529. circuit to an exit that may be able to connect to the destination
  530. address, selects an arbitrary StreamID not yet used on that circuit,
  531. and constructs a RELAY_BEGIN cell with a payload encoding the address
  532. and port of the destination host. The payload format is:
  533. ADDRESS | ':' | PORT | [00]
  534. where ADDRESS can be a DNS hostname, or an IPv4 address in
  535. dotted-quad format, or an IPv6 address surrounded by square brackets;
  536. and where PORT is encoded in decimal.
  537. [What is the [00] for? -NM]
  538. [It's so the payload is easy to parse out with string funcs -RD]
  539. Upon receiving this cell, the exit node resolves the address as
  540. necessary, and opens a new TCP connection to the target port. If the
  541. address cannot be resolved, or a connection can't be established, the
  542. exit node replies with a RELAY_END cell. (See 6.4 below.)
  543. Otherwise, the exit node replies with a RELAY_CONNECTED cell, whose
  544. payload is in one of the following formats:
  545. The IPv4 address to which the connection was made [4 octets]
  546. A number of seconds (TTL) for which the address may be cached [4 octets]
  547. or
  548. Four zero-valued octets [4 octets]
  549. An address type (6) [1 octet]
  550. The IPv6 address to which the connection was made [16 octets]
  551. A number of seconds (TTL) for which the address may be cached [4 octets]
  552. [XXXX Versions of Tor before 0.1.1.6 ignore and do not generate the TTL
  553. field. No version of Tor currently generates the IPv6 format.
  554. Tor servers before 0.1.2.0 set the TTL field to a fixed value. Later
  555. versions set the TTL to the last value seen from a DNS server, and expire
  556. their own cached entries after a fixed interval. This prevents certain
  557. attacks.]
  558. The OP waits for a RELAY_CONNECTED cell before sending any data.
  559. Once a connection has been established, the OP and exit node
  560. package stream data in RELAY_DATA cells, and upon receiving such
  561. cells, echo their contents to the corresponding TCP stream.
  562. RELAY_DATA cells sent to unrecognized streams are dropped.
  563. Relay RELAY_DROP cells are long-range dummies; upon receiving such
  564. a cell, the OR or OP must drop it.
  565. 6.2.1. Opening a directory stream
  566. If a Tor server is a directory server, it should respond to a
  567. RELAY_BEGIN_DIR cell as if it had received a BEGIN cell requesting a
  568. connection to its directory port. RELAY_BEGIN_DIR cells ignore exit
  569. policy, since the stream is local to the Tor process.
  570. If the Tor server is not running a directory service, it should respond
  571. with a REASON_NOTDIRECTORY RELAY_END cell.
  572. Clients MUST generate an all-zero payload for RELAY_BEGIN_DIR cells,
  573. and servers MUST ignore the payload.
  574. [RELAY_BEGIN_DIR was not supported before Tor 0.1.2.2-alpha; clients
  575. SHOULD NOT send it to routers running earlier versions of Tor.]
  576. 6.3. Closing streams
  577. When an anonymized TCP connection is closed, or an edge node
  578. encounters error on any stream, it sends a 'RELAY_END' cell along the
  579. circuit (if possible) and closes the TCP connection immediately. If
  580. an edge node receives a 'RELAY_END' cell for any stream, it closes
  581. the TCP connection completely, and sends nothing more along the
  582. circuit for that stream.
  583. The payload of a RELAY_END cell begins with a single 'reason' byte to
  584. describe why the stream is closing, plus optional data (depending on
  585. the reason.) The values are:
  586. 1 -- REASON_MISC (catch-all for unlisted reasons)
  587. 2 -- REASON_RESOLVEFAILED (couldn't look up hostname)
  588. 3 -- REASON_CONNECTREFUSED (remote host refused connection) [*]
  589. 4 -- REASON_EXITPOLICY (OR refuses to connect to host or port)
  590. 5 -- REASON_DESTROY (Circuit is being destroyed)
  591. 6 -- REASON_DONE (Anonymized TCP connection was closed)
  592. 7 -- REASON_TIMEOUT (Connection timed out, or OR timed out
  593. while connecting)
  594. 8 -- (unallocated) [**]
  595. 9 -- REASON_HIBERNATING (OR is temporarily hibernating)
  596. 10 -- REASON_INTERNAL (Internal error at the OR)
  597. 11 -- REASON_RESOURCELIMIT (OR has no resources to fulfill request)
  598. 12 -- REASON_CONNRESET (Connection was unexpectedly reset)
  599. 13 -- REASON_TORPROTOCOL (Sent when closing connection because of
  600. Tor protocol violations.)
  601. 14 -- REASON_NOTDIRECTORY (Client sent RELAY_BEGIN_DIR to a
  602. non-directory server.)
  603. (With REASON_EXITPOLICY, the 4-byte IPv4 address or 16-byte IPv6 address
  604. forms the optional data; no other reason currently has extra data.
  605. As of 0.1.1.6, the body also contains a 4-byte TTL.)
  606. OPs and ORs MUST accept reasons not on the above list, since future
  607. versions of Tor may provide more fine-grained reasons.
  608. [*] Older versions of Tor also send this reason when connections are
  609. reset.
  610. [**] Due to a bug in versions of Tor through 0095, error reason 8 must
  611. remain allocated until that version is obsolete.
  612. --- [The rest of this section describes unimplemented functionality.]
  613. Because TCP connections can be half-open, we follow an equivalent
  614. to TCP's FIN/FIN-ACK/ACK protocol to close streams.
  615. An exit connection can have a TCP stream in one of three states:
  616. 'OPEN', 'DONE_PACKAGING', and 'DONE_DELIVERING'. For the purposes
  617. of modeling transitions, we treat 'CLOSED' as a fourth state,
  618. although connections in this state are not, in fact, tracked by the
  619. onion router.
  620. A stream begins in the 'OPEN' state. Upon receiving a 'FIN' from
  621. the corresponding TCP connection, the edge node sends a 'RELAY_FIN'
  622. cell along the circuit and changes its state to 'DONE_PACKAGING'.
  623. Upon receiving a 'RELAY_FIN' cell, an edge node sends a 'FIN' to
  624. the corresponding TCP connection (e.g., by calling
  625. shutdown(SHUT_WR)) and changing its state to 'DONE_DELIVERING'.
  626. When a stream in already in 'DONE_DELIVERING' receives a 'FIN', it
  627. also sends a 'RELAY_FIN' along the circuit, and changes its state
  628. to 'CLOSED'. When a stream already in 'DONE_PACKAGING' receives a
  629. 'RELAY_FIN' cell, it sends a 'FIN' and changes its state to
  630. 'CLOSED'.
  631. If an edge node encounters an error on any stream, it sends a
  632. 'RELAY_END' cell (if possible) and closes the stream immediately.
  633. 6.4. Remote hostname lookup
  634. To find the address associated with a hostname, the OP sends a
  635. RELAY_RESOLVE cell containing the hostname to be resolved. (For a reverse
  636. lookup, the OP sends a RELAY_RESOLVE cell containing an in-addr.arpa
  637. address.) The OR replies with a RELAY_RESOLVED cell containing a status
  638. byte, and any number of answers. Each answer is of the form:
  639. Type (1 octet)
  640. Length (1 octet)
  641. Value (variable-width)
  642. TTL (4 octets)
  643. "Length" is the length of the Value field.
  644. "Type" is one of:
  645. 0x00 -- Hostname
  646. 0x04 -- IPv4 address
  647. 0x06 -- IPv6 address
  648. 0xF0 -- Error, transient
  649. 0xF1 -- Error, nontransient
  650. If any answer has a type of 'Error', then no other answer may be given.
  651. The RELAY_RESOLVE cell must use a nonzero, distinct streamID; the
  652. corresponding RELAY_RESOLVED cell must use the same streamID. No stream
  653. is actually created by the OR when resolving the name.
  654. 7. Flow control
  655. 7.1. Link throttling
  656. Each node should do appropriate bandwidth throttling to keep its
  657. user happy.
  658. Communicants rely on TCP's default flow control to push back when they
  659. stop reading.
  660. 7.2. Link padding
  661. Link padding can be created by sending PADDING cells along the
  662. connection; relay cells of type "DROP" can be used for long-range
  663. padding.
  664. Currently nodes are not required to do any sort of link padding or
  665. dummy traffic. Because strong attacks exist even with link padding,
  666. and because link padding greatly increases the bandwidth requirements
  667. for running a node, we plan to leave out link padding until this
  668. tradeoff is better understood.
  669. 7.3. Circuit-level flow control
  670. To control a circuit's bandwidth usage, each OR keeps track of
  671. two 'windows', consisting of how many RELAY_DATA cells it is
  672. allowed to package for transmission, and how many RELAY_DATA cells
  673. it is willing to deliver to streams outside the network.
  674. Each 'window' value is initially set to 1000 data cells
  675. in each direction (cells that are not data cells do not affect
  676. the window). When an OR is willing to deliver more cells, it sends a
  677. RELAY_SENDME cell towards the OP, with Stream ID zero. When an OR
  678. receives a RELAY_SENDME cell with stream ID zero, it increments its
  679. packaging window.
  680. Each of these cells increments the corresponding window by 100.
  681. The OP behaves identically, except that it must track a packaging
  682. window and a delivery window for every OR in the circuit.
  683. An OR or OP sends cells to increment its delivery window when the
  684. corresponding window value falls under some threshold (900).
  685. If a packaging window reaches 0, the OR or OP stops reading from
  686. TCP connections for all streams on the corresponding circuit, and
  687. sends no more RELAY_DATA cells until receiving a RELAY_SENDME cell.
  688. [this stuff is badly worded; copy in the tor-design section -RD]
  689. 7.4. Stream-level flow control
  690. Edge nodes use RELAY_SENDME cells to implement end-to-end flow
  691. control for individual connections across circuits. Similarly to
  692. circuit-level flow control, edge nodes begin with a window of cells
  693. (500) per stream, and increment the window by a fixed value (50)
  694. upon receiving a RELAY_SENDME cell. Edge nodes initiate RELAY_SENDME
  695. cells when both a) the window is <= 450, and b) there are less than
  696. ten cell payloads remaining to be flushed at that edge.
  697. A.1. Differences between spec and implementation
  698. - The current specification requires all ORs to have IPv4 addresses, but
  699. allows servers to exit and resolve to IPv6 addresses, and to declare IPv6
  700. addresses in their exit policies. The current codebase has no IPv6
  701. support at all.
  702. B. Things that should change in a later version of the Tor protocol
  703. B.1. ... but which will require backward-incompatible change
  704. - Circuit IDs should be longer.
  705. - IPv6 everywhere.
  706. - Maybe, keys should be longer.
  707. - Maybe, key-length should be adjustable. How to do this without
  708. making anonymity suck?
  709. - Drop backward compatibility.
  710. - We should use a 128-bit subgroup of our DH prime.
  711. - Handshake should use HMAC.
  712. - Multiple cell lengths.
  713. - Ability to split circuits across paths (If this is useful.)
  714. - SENDME windows should be dynamic.
  715. - Directory
  716. - Stop ever mentioning socks ports
  717. B.1. ... and that will require no changes
  718. - Mention multiple addr/port combos
  719. - Advertised outbound IP?
  720. - Migrate streams across circuits.
  721. B.2. ... and that we have no idea how to do.
  722. - UDP (as transport)
  723. - UDP (as content)
  724. - Use a better AES mode that has built-in integrity checking,
  725. doesn't grow with the number of hops, is not patented, and
  726. is implemented and maintained by smart people.