tor-dev September 2014

tor-dev@lists.torproject.org

57 participants
45 discussions

(Draft) Proposal 224: Next-Generation Hidden Services in Tor
by Nick Mathewson 28 May '15

28 May '15

Hi, all! I've been trying to fill in all the cracks and corners for a revamp of the hidden services protocol, based on earlier writings by George Kadianakis and other discussions on the mailing list. (See draft acknowledgments section below.) After a bunch of comments, I'm ready to give this a number and call it (draft) proposal 224. I'd like to know what doesn't make sense, what I need to explain better, and what I need to design better. I'd like to fill in the gaps and turn this into a more full document. I'd like to answer the open questions. Comments are most welcome, especially if they grow into improvements. FWIW, I am likely to be offline for most of the current weekend, because of Thanksgiving, so please be patient with my reply speed; I hope to catch up with emails next week. Filename: 224-rend-spec-ng.txt Title: Next-Generation Hidden Services in Tor Author: Nick Mathewson Created: 2013-11-29 Status: Draft -1. Draft notes This document describes a proposed design and specification for hidden services in Tor version 0.2.5.x or later. It's a replacement for the current rend-spec.txt, rewritten for clarity and for improved design. Look for the string "TODO" below: it describes gaps or uncertainties in the design. Change history: 2013-11-29: Proposal first numbered. Some TODO and XXX items remain. 0. Hidden services: overview and preliminaries. Hidden services aim to provide responder anonymity for bidirectional stream-based communication on the Tor network. Unlike regular Tor connections, where the connection initiator receives anonymity but the responder does not, hidden services attempt to provide bidirectional anonymity. Other features include: * [TODO: WRITE ME once there have been some more drafts and we know what the summary should say.] Participants: Operator -- A person running a hidden service Host, "Server" -- The Tor software run by the operator to provide a hidden service. User -- A person contacting a hidden service. Client -- The Tor software running on the User's computer Hidden Service Directory (HSDir) -- A Tor node that hosts signed statements from hidden service hosts so that users can make contact with them. Introduction Point -- A Tor node that accepts connection requests for hidden services and anonymously relays those requests to the hidden service. Rendezvous Point -- A Tor node to which clients and servers connect and which relays traffic between them. 0.1. Improvements over previous versions. [TODO write me once there have been more drafts and we know what the summary should say.] 0.2. Notation and vocabulary Unless specified otherwise, all multi-octet integers are big-endian. We write sequences of bytes in two ways: 1. A sequence of two-digit hexadecimal values in square brackets, as in [AB AD 1D EA]. 2. A string of characters enclosed in quotes, as in "Hello". These characters in these string are encoded in their ascii representations; strings are NOT nul-terminated unless explicitly described as NUL terminated. We use the words "byte" and "octet" interchangeably. We use the vertical bar | to denote concatenation. We use INT_N(val) to denote the network (big-endian) encoding of the unsigned integer "val" in N bytes. For example, INT_4(1337) is [00 00 05 39]. 0.3. Cryptographic building blocks This specification uses the following cryptographic building blocks: * A stream cipher STREAM(iv, k) where iv is a nonce of length S_IV_LEN bytes and k is a key of length S_KEY_LEN bytes. * A public key signature system SIGN_KEYGEN()->seckey, pubkey; SIGN_SIGN(seckey,msg)->sig; and SIGN_CHECK(pubkey, sig, msg) -> { "OK", "BAD" }; where secret keys are of length SIGN_SECKEY_LEN bytes, public keys are of length SIGN_PUBKEY_LEN bytes, and signatures are of length SIGN_SIG_LEN bytes. This signature system must also support key blinding operations as discussed in appendix [KEYBLIND] and in section [SUBCRED]: SIGN_BLIND_SECKEY(seckey, blind)->seckey2 and SIGN_BLIND_PUBKEY(pubkey, blind)->pubkey2 . * A public key agreement system "PK", providing PK_KEYGEN()->seckey, pubkey; PK_VALID(pubkey) -> {"OK", "BAD"}; and PK_HANDHAKE(seckey, pubkey)->output; where secret keys are of length PK_SECKEY_LEN bytes, public keys are of length PK_PUBKEY_LEN bytes, and the handshake produces outputs of length PK_OUTPUT_LEN bytes. * A cryptographic hash function H(d), which should be preimage and collision resistant. It produces hashes of length HASH_LEN bytes. * A cryptographic message authentication code MAC(key,msg) that produces outputs of length MAC_LEN bytes. * A key derivation function KDF(key data, salt, personalization, n) that outputs n bytes. As a first pass, I suggest: * Instantiate STREAM with AES128-CTR. [TODO: or ChaCha20?] * Instantiate SIGN with Ed25519 and the blinding protocol in [KEYBLIND]. * Instantiate PK with Curve25519. * Instantiate H with SHA256. [TODO: really?] * Instantiate MAC with HMAC using H. * Instantiate KDF with HKDF using H. For legacy purposes, we specify compatibility with older versions of the Tor introduction point and rendezvous point protocols. These used RSA1024, DH1024, AES128, and SHA1, as discussed in rend-spec.txt. Except as noted, all RSA keys MUST have exponent values of 65537. As in [proposal 220], all signatures are generated not over strings themselves, but over those strings prefixed with a distinguishing value. 0.4. Protocol building blocks [BUILDING-BLOCKS] In sections below, we need to transmit the locations and identities of Tor nodes. We do so in the link identification format used by EXTEND2 cells in the Tor protocol. NSPEC (Number of link specifiers) [1 byte] NSPEC times: LSTYPE (Link specifier type) [1 byte] LSLEN (Link specifier length) [1 byte] LSPEC (Link specifier) [LSLEN bytes] Link specifier types are as described in tor-spec.txt. Every set of link specifiers MUST include at minimum specifiers of type [00] (TLS-over-TCP, IPv4) and [02] (legacy node identity). We also incorporate Tor's circuit extension handshakes, as used in the CREATE2 and CREATED2 cells described in tor-spec.txt. In these handshakes, a client who knows a public key for a server sends a message and receives a message from that server. Once the exchange is done, the two parties have a shared set of forward-secure key material, and the client knows that nobody else shares that key material unless they control the secret key corresponding to the server's public key. 0.5. Assigned relay cell types These relay cell types are reserved for use in the hidden service protocol. 32 -- RELAY_COMMAND_ESTABLISH_INTRO Sent from hidden service host to introduction point; establishes introduction point. Discussed in [REG_INTRO_POINT]. 33 -- RELAY_COMMAND_ESTABLISH_RENDEZVOUS Sent from client to rendezvous point; creates rendezvous point. Discussed in [EST_REND_POINT]. 34 -- RELAY_COMMAND_INTRODUCE1 Sent from client to introduction point; requests introduction. Discussed in [SEND_INTRO1] 35 -- RELAY_COMMAND_INTRODUCE2 Sent from client to introduction point; requests introduction. Same format as INTRODUCE1. Discussed in [FMT_INTRO1] and [PROCESS_INTRO2] 36 -- RELAY_COMMAND_RENDEZVOUS1 Sent from introduction point to rendezvous point; attempts to join introduction point's circuit to client's circuit. Discussed in [JOIN_REND] 37 -- RELAY_COMMAND_RENDEZVOUS2 Sent from introduction point to rendezvous point; reports join of introduction point's circuit to client's circuit. Discussed in [JOIN_REND] 38 -- RELAY_COMMAND_INTRO_ESTABLISHED Sent from introduction point to hidden service host; reports status of attempt to establish introduction point. Discussed in [INTRO_ESTABLISHED] 39 -- RELAY_COMMAND_RENDEZVOUS_ESTABLISHED Sent from rendezvous point to client; acknowledges receipt of ESTABLISH_RENDEZVOUS cell. Discussed in [EST_REND_POINT] 40 -- RELAY_COMMAND_INTRODUCE_ACK Sent form introduction point to client; acknowledges receipt of INTRODUCE1 cell and reports success/failure. Discussed in [INTRO_ACK] 0.5. Acknowledgments [TODO reformat these once the lists are more complete.] This design includes ideas from many people, including Christopher Baines, Daniel J. Bernstein, Matthew Finkel, Ian Goldberg, George Kadianakis, Aniket Kate, Tanja Lange, Robert Ransom, It's based on Tor's original hidden service design by Roger Dingledine, Nick Mathewson, and Paul Syverson, and on improvements to that design over the years by people including Tobias Kamm, Thomas Lauterbach, Karsten Loesing, Alessandro Preite Martinez, Robert Ransom, Ferdinand Rieger, Christoph Weingarten, Christian Wilms, We wouldn't be able to do any of this work without good attack designs from researchers including Alex Biryukov, Lasse Øverlier, Ivan Pustogarov, Paul Syverson Ralf-Philipp Weinmann, See [ATTACK-REFS] for their papers. Several of these ideas have come from conversations with Christian Grothoff, Brian Warner, Zooko Wilcox-O'Hearn, And if this document makes any sense at all, it's thanks to editing help from Matthew Finkel George Kadianakis, Peter Palfrader, [XXX Acknowledge the huge bunch of people working on 8106.] [XXX Acknowledge the huge bunch of people working on 8244.] Please forgive me if I've missed you; please forgive me if I've misunderstood your best ideas here too. 1. Protocol overview In this section, we outline the hidden service protocol. This section omits some details in the name of simplicity; those are given more fully below, when we specify the protocol in more detail. 1.1. View from 10,000 feet A hidden service host prepares to offer a hidden service by choosing several Tor nodes to serve as its introduction points. It builds circuits to those nodes, and tells them to forward introduction requests to it using those circuits. Once introduction points have been picked, the host builds a set of documents called "hidden service descriptors" (or just "descriptors" for short) and uploads them to a set of HSDir nodes. These documents list the hidden service's current introduction points and describe how to make contact with the hidden service. When a client wants to connect to a hidden service, it first chooses a Tor node at random to be its "rendezvous point" and builds a circuit to that rendezvous point. If the client does not have an up-to-date descriptor for the service, it contacts an appropriate HSDir and requests such a descriptor. The client then builds an anonymous circuit to one of the hidden service's introduction points listed in its descriptor, and gives the introduction point an introduction request to pass to the hidden service. This introduction request includes the target rendezvous point and the first part of a cryptographic handshake. Upon receiving the introduction request, the hidden service host makes an anonymous circuit to the rendezvous point and completes the cryptographic handshake. The rendezvous point connects the two circuits, and the cryptographic handshake gives the two parties a shared key and proves to the client that it is indeed talking to the hidden service. Once the two circuits are joined, the client can send Tor RELAY cells to the server. RELAY_BEGIN cells open streams to an external process or processes configured by the server; RELAY_DATA cells are used to communicate data on those streams, and so forth. 1.2. In more detail: naming hidden services [NAMING] A hidden service's name is its long term master identity key. This is encoded as a hostname by encoding the entire key in Base 32, and adding the string ".onion" at the end. (This is a change from older versions of the hidden service protocol, where we used an 80-bit truncated SHA1 hash of a 1024 bit RSA key.) The names in this format are distinct from earlier names because of their length. An older name might look like: unlikelynamefora.onion yyhws9optuwiwsns.onion And a new name following this specification might look like: a1uik0w1gmfq3i5ievxdm9ceu27e88g6o7pe0rffdw9jmntwkdsd.onion Note that since master keys are 32 bytes long, and 52 bytes of base 32 encoding can hold 260 bits of information, we have four unused bits in each of these names. [TODO: Alternatively, we could require that the first bit of the master key always be zero, and use a 51-byte encoding. Or we could require that the first two bits be zero, and use a 51-byte encoding and reserve the first bit. Or we could require that the first nine bits, or ten bits be zero, etc.] 1.3. In more detail: Access control [IMD:AC] Access control for a hidden service is imposed at multiple points through the process above. In order to download a descriptor, clients must know which blinded signing key was used to sign it. (See the next section for more info on key blinding.) This blinded signing key is derived from the service's public key and, optionally, an additional secret that is not part of the hidden service's onion address. The public key and this secret together constitute the service's "credential". When the secret is in use, the hidden service gains protections equivalent to the "stealth mode" in previous designs. To learn the introduction points, the clients must decrypt the body of the hidden service descriptor. The encryption key for these is derived from the service's credential. In order to make an introduction point send a request to the server, the client must know the introduction point and know the service's per-introduction-point authentication key from the hidden service descriptor. The final level of access control happens at the server itself, which may decide to respond or not respond to the client's request depending on the contents of the request. The protocol is extensible at this point: at a minimum, the server requires that the client demonstrate knowledge od the contents of the encrypted portion of the hidden service descriptor. The service may additionally require a user- or group-specific access token before it responds to requests. 1.4. In more detail: Distributing hidden service descriptors. [IMD:DIST] Periodically, hidden service descriptors become stored at different locations to prevent a single directory or small set of directories from becoming a good DoS target for removing a hidden service. For each period, the Tor directory authorities agree upon a collaboratively generated random value. (See section 2.3 for a description of how to incorporate this value into the voting practice; generating the value is described in other proposals, including [TODO: add a reference]) That value, combined with hidden service directories' public identity keys, determines each HSDirs' position in the hash ring for descriptors made in that period. Each hidden service's descriptors are placed into the ring in positions based on the key that was used to sign them. Note that hidden service descriptors are not signed with the services' public keys directly. Instead, we use a key-blinding system [KEYBLIND] to create a new key-of-the-day for each hidden service. Any client that knows the hidden service's credential can derive these blinded signing keys for a given period. It should be impossible to derive the blinded signing key lacking that credential. The body of each descriptor is also encrypted with a key derived from the credential. To avoid a "thundering herd" problem where every service generates and uploads a new descriptor at the start of each period, each descriptor comes online at a time during the period that depends on its blinded signing key. The keys for the last period remain valid until the new keys come online. 1.5. In more detail: Scaling to multiple hosts [THIS SECTION IS UNFINISHED] In order to allow multiple hosts to provide a single hidden service, I'm considering two options. * We can have each server build an introduction circuit to each introduction point, and have the introduction points responsible for round-robining between these circuits. One service host is responsible for picking the introduction points and publishing the descriptors. * We can have servers choose their introduction points independently, and build circuits to them. One service host is responsible for combining these introduction points into a single descriptor. If we want to avoid having a single "master" host without which the whole service goes down (the "one service host" in the description above), we need a way to fail over from one host to another. We also need a way to coordinate between the hosts. This is as yet undesigned. Maybe it should use a hidden service? See [SCALING-REFS] for discussion on this topic. [TODO: Finalize this design.] 1.6. In more detail: Backward compatibility with older hidden service protocols This design is incompatible with the clients, server, and hsdir node protocols from older versions of the hidden service protocol as described in rend-spec.txt. On the other hand, it is designed to enable the use of older Tor nodes as rendezvous points and introduction points. 1.7. In more detail: Offline operation In this design, a hidden service's secret identity key may be stored offline. It's used only to generate blinded identity keys, which are used to sign descriptor signing keys. In order to operate a hidden service, the operator can generate a number of descriptor signing keys and their certifications (see [DESC-OUTER] and [ENCRYPTED-DATA] below), and their corresponding descriptor encryption keys, and export those to the hidden service hosts. 1.8. In more detail: Encryption Keys And Replay Resistance To avoid replays of an introduction request by an introduction point, a hidden service host must never accept the same request twice. Earlier versions of the hidden service design used a authenticated timestamp here, but including a view of the current time can create a problematic fingerprint. (See proposal 222 for more discussion.) 1.9. In more detail: A menagerie of keys [In the text below, an "encryption keypair" is roughly "a keypair you can do Diffie-Hellman with" and a "signing keypair" is roughly "a keypair you can do ECDSA with."] Public/private keypairs defined in this document: Master (hidden service) identity key -- A master signing keypair used as the identity for a hidden service. This key is not used on its own to sign anything; it is only used to generate blinded signing keys as described in [KEYBLIND] and [SUBCRED]. Blinded signing key -- A keypair derived from the identity key, used to sign descriptor signing keys. Changes periodically for each service. Clients who know a 'credential' consisting of the service's public identity key and an optional secret can derive the public blinded identity key for a service. This key is used as an index in the DHT-like structure of the directory system. Descriptor signing key -- A key used to sign hidden service descriptors. This is signed by blinded signing keys. Unlike blinded signing keys and master identity keys, the secret part of this key must be stored online by hidden service hosts. Introduction point authentication key -- A short-term signing keypair used to identify a hidden service to a given introduction point. A fresh keypair is made for each introduction point; these are used to sign the request that a hidden service host makes when establishing an introduction point, so that clients who know the public component of this key can get their introduction requests sent to the right service. No keypair is ever used with more than one introduction point. (previously called a "service key" in rend-spec.txt) Introduction point encryption key -- A short-term encryption keypair used when establishing connections via an introduction point. Plays a role analogous to Tor nodes' onion keys. A fresh keypair is made for each introduction point. Symmetric keys defined in this document: Descriptor encryption keys -- A symmetric encryption key used to encrypt the body of hidden service descriptors. Derived from the current period and the hidden service credential. Public/private keypairs defined elsewhere: Onion key -- Short-term encryption keypair (Node) identity key Symmetric key-like things defined elsewhere: KH from circuit handshake -- An unpredictable value derived as part of the Tor circuit extension handshake, used to tie a request to a particular circuit. 2. Generating and publishing hidden service descriptors [HSDIR] Hidden service descriptors follow the same metaformat as other Tor directory objects. They are published anonymously to Tor servers with the HSDir3 flag. (Authorities should assign this flag as they currently assign the HSDir flag, except that they should restrict it to Tor versions implementing the HSDir parts of this specification.) 2.1. Deriving blinded keys and subcredentials [SUBCRED] In each time period (see [TIME-PERIOD] for a definition of time periods), a hidden service host uses a different blinded private key to sign its directory information, and clients use a different blinded public key as the index for fetching that information. For a candidate for a key derivation method, see Appendix [KEYBLIND]. Additionally, clients and hosts derive a subcredential for each period. Knowledge of the subcredential is needed to decrypt hidden service descriptors for each period and to authenticate with the hidden service host in the introduction process. Unlike the credential, it changes each period. Knowing the subcredential, even in combination with the blinded private key, does not enable the hidden service host to derive the main credential--therefore, it is safe to put the subcredential on the hidden service host while leaving the hidden service's private key offline. The subcredential for a period is derived as: H("subcredential" | credential | blinded-public-key). 2.2. Locating, uploading, and downloading hidden service descriptors [HASHRING] To avoid attacks where a hidden service's descriptor is easily targeted for censorship, we store them at different directories over time, and use shared random values to prevent those directories from being predictable far in advance. Which Tor servers hosts a hidden service depends on: * the current time period, * the daily subcredential, * the hidden service directories' public keys, * a shared random value that changes in each time period, * a set of network-wide networkstatus consensus parameters. Below we explain in more detail. 2.2.1. Dividing time into periods [TIME-PERIODS] To prevent a single set of hidden service directory from becoming a target by adversaries looking to permanently censor a hidden service, hidden service descriptors are uploaded to different locations that change over time. The length of a "time period" is controlled by the consensus parameter 'hsdir-interval', and is a number of minutes between 30 and 14400 (10 days). The default time period length is 1500 (one day plus one hour). Time periods start with the Unix epoch (Jan 1, 1970), and are computed by taking the number of whole minutes since the epoch and dividing by the time period. So if the current time is 2013-11-12 13:44:32 UTC, making the seconds since the epoch 1384281872, the number of minutes since the epoch is 23071364. If the current time period length is 1500 (the default), then the current time period number is 15380. It began 15380*1500*60 seconds after the epoch at 2013-11-11 20:00:00 UTC, and will end at (15380+1)*1500*60 seconds after the epoch at 2013-11-12 21:00:00 UTC. 2.2.2. Overlapping time periods to avoid thundering herds [TIME-OVERLAP] If every hidden service host were to generate a new set of keys and upload a new descriptor at exactly the start of each time period, the directories would be overwhelmed by every host uploading at the same time. Instead, each public key becomes valid at its new location at a deterministic time somewhat _before_ the period begins, depending on the public key and the period. The time at which a key might first become valid is determined by the consensus parameter "hsdir-overlap-begins", which is an integer in range [1,100] with default value 80. This parameter denotes a percentage of the interval for which no overlap occurs. So for the default interval (1500 minutes) and default overlap-begins value (80%), new keys do not become valid for the first 1200 minutes of the interval. The new shared random value must be published *before* the start of the next overlap interval by at least enough time to ensure that clients all get it. [TODO: how much earlier?] The time at which a key from the next interval becomes valid is determined by taking the first two bytes of OFFSET = H(Key | INT_8(Next_Period_Num)) as a big-endian integer, dividing by 65536, and treating that as a fraction of the overlap interval. For example, if the period is 1500 minutes long, and overlap interval is 300 minutes long, and OFFSET begins with [90 50], then the next key becomes valid at 1200 + 300 * (0x9050 / 65536) minutes, or approximately 22 hours and 49 minutes after the beginning of the period. Hidden service directories should accept descriptors at least [TODO: how much?] minutes before they would become valid, and retain them for at least [TODO: how much?] minutes after the end of the period. When a client is looking for a service, it must calculate its key both for the current and for the subsequent period, to decide whether the next period's key is valid yet. 2.2.3. Where to publish a service descriptor The following consensus parameters control where a hidden service descriptor is stored; hsdir_n_replicas = an integer in range [1,16] with default value 2. hsdir_spread_fetch = an integer in range [1,128] with default value 3. hsdir_spread_store = an integer in range [1,128] with default value 3. hsdir_spread_accept = an integer in range [1,128] with default value 8. To determine where a given hidden service descriptor will be stored in a given period, after the blinded public key for that period is derived, the uploading or downloading party calculate for replicanum in 1...hsdir_n_replicas: hs_index(replicanum) = H("store-at-idx" | blinded_public_key | replicanum | periodnum) where blinded_public_key is specified in section KEYBLIND, and periodnum is defined in section TIME-PERIODS. where n_replicas is determined by the consensus parameter "hsdir_n_replicas". Then, for each node listed in the current consensus with the HSDir3 flag, we compute a directory index for that node as: hsdir_index(node) = H(node_identity_digest | shared_random | INT_8(period_num) ) where shared_random is the shared value generated by the authorities in section PUB-SHAREDRANDOM. Finally, for replicanum in 1...hsdir_n_replicas, the hidden service host uploads descriptors to the first hsdir_spread_store nodes whose indices immediately follow hs_index(replicanum). When choosing an HSDir to download from, clients choose randomly from among the first hsdir_spread_fetch nodes after the indices. (Note that, in order to make the system better tolerate disappearing HSDirs, hsdir_spread_fetch may be less than hsdir_spread_store.) An HSDir should rejects a descriptor if that HSDir is not one of the first hsdir_spread_accept HSDirs for that node. [TODO: Incorporate the findings from proposal 143 here. But watch out: proposal 143 did not analyze how much the set of nodes changes over time, or how much client and host knowledge might diverge.] 2.2.4. URLs for anonymous uploading and downloading Hidden service descriptors conforming to this specification are uploaded with an HTTP POST request to the URL /tor/rendezvous3/publish relative to the hidden service directory's root, and downloaded with an HTTP GET request for the URL /tor/rendezvous3/<z> where z is a base-64 encoding of the hidden service's blinded public key. [TODO: raw base64 is not super-nice for URLs, since it can have slashes. We already use it for microdescriptor URLs, though. Do we care here?] These requests must be made anonymously, on circuits not used for anything else. 2.3. Publishing shared random values [PUB-SHAREDRANDOM] Our design for limiting the predictability of HSDir upload locations relies on a shared random value that isn't predictable in advance or too influenceable by an attacker. The authorities must run a protocol to generate such a value at least once per hsdir period. Here we describe how they publish these values; the procedure they use to generate them can change independently of the rest of this specification. For one possible (somewhat broken) protocol, see Appendix [SHAREDRANDOM]. We add a new line in votes and consensus documents: "hsdir-shared-random" PERIOD-START VALUE PERIOD-START = YYYY-MM-DD HH:MM:SS VALUE = A base-64 encoded 256-bit value. To decide which hsdir-shared-random line to include in a consensus for a given PERIOD-START, we choose whichever line appears verbatim in the most votes, so long as it is listed by at least three authorities. Ties are broken in favor of the lower value. More than one PERIOD-START is allowed per vote, and per consensus. The same PERIOD-START must not appear twice in a vote or in a consensus. [TODO: Need to define a more robust algorithm. Need to cover cases where multiple cluster of authorities publish a different value, etc.] The hs-dir-shared-random lines appear, sorted by PERIOD-START, in the consensus immediately after the "params" line. The authorities should publish the shared random value for the current period, and, at a time at least three voting periods before the overlap interval begins, the shared random value for the next period. [TODO: find out what weasel doesn't like here.] 2.4. Hidden service descriptors: outer wrapper [DESC-OUTER] The format for a hidden service descriptor is as follows, using the meta-format from dir-spec.txt. "hs-descriptor" SP "3" SP public-key SP certification NL [At start, exactly once.] public-key is the blinded public key for the service, encoded in base 64. Certification is a certification of a short-term ed25519 descriptor signing key using the public key, in the format of proposal 220. "time-period" SP YYYY-MM-DD HH:MM:SS NUM NL [Exactly once.] The time period for which this descriptor is relevant, including its starting time and its period number. "revision-counter" SP Integer NL [Exactly once.] The revision number of the descriptor. If an HSDir receives a second descriptor for a key that it already has a descriptor for, it should retain and serve the descriptor with the higher revision-counter. (Checking for monotonically increasing revision-counter values prevents an attacker from replacing a newer descriptor signed by a given key with a copy of an older version.) "encrypted" NL encrypted-string [Exactly once.] An encrypted blob, whose format is discussed in [ENCRYPTED-DATA] below. The blob is base-64 encoded and enclosed in -----BEGIN MESSAGE---- and ----END MESSAGE---- wrappers. "signature" SP signature NL [exactly once, at end.] A signature of all previous fields, using the signing key in the hs-descriptor line. We use a separate key for signing, so that the hidden service host does not need to have its private blinded key online. 2.5. Hidden service descriptors: encryption format [ENCRYPTED-DATA] The encrypted part of the hidden service descriptor is encrypted and authenticated with symmetric keys generated as follows: salt = 16 random bytes secret_input = nonce | blinded_public_key | subcredential | INT_4(revision_counter) keys = KDF(secret_input, salt, "hsdir-encrypted-data", S_KEY_LEN + S_IV_LEN + MAC_KEY_LEN) SECRET_KEY = first S_KEY_LEN bytes of keys SECRET_IV = next S_IV_LEN bytes of keys MAC_KEY = last MAC_KEY_LEN bytes of keys The encrypted data has the format: SALT (random bytes from above) [16 bytes] ENCRYPTED The plaintext encrypted with S [variable] MAC MAC of both above fields [32 bytes] The encryption format is ENCRYPTED = STREAM(SECRET_IV,SECRET_KEY) xor Plaintext Before encryption, the plaintext must be padded to a multiple of ??? bytes with NUL bytes. The plaintext must not be longer than ??? bytes. [TODO: how much? Should this be a parameter? What values in practice is needed to hide how many intro points we have, and how many might be legacy ones?] The plaintext format is: "create2-formats" SP formats NL [Exactly once] A space-separated list of integers denoting CREATE2 cell format numbers that the server recognizes. Must include at least TAP and ntor as described in tor-spec.txt. See tor-spec section 5.1 for a list of recognized handshake types. "authentication-required" SP types NL [At most once] A space-separated list of authentication types. A client that does not support at least one of these authentication types will not be able to contact the host. Recognized types are: 'password' and 'ed25519'. See [INTRO-AUTH] below. At least once: "introduction-point" SP link-specifiers NL [Exactly once per introduction point at start of introduction point section] The link-specifiers is a base64 encoding of a link specifier block in the format described in BUILDING-BLOCKS. "auth-key" SP "ed25519" SP key SP certification NL [Exactly once per introduction point] Base-64 encoded introduction point authentication key that was used to establish introduction point circuit, cross-certifying the blinded public key key using the certification format of proposal 220. "enc-key" SP "ntor" SP key NL [At most once per introduction point] Base64-encoded curve25519 key used to encrypt request to hidden service. [TODO: I'd like to have a cross-certification here too.] "enc-key" SP "legacy" NL key NL [At most once per introduction point] Base64-encoded RSA key, wrapped in "----BEGIN RSA PUBLIC KEY-----" armor, for use with a legacy introduction point as described in [LEGACY_EST_INTRO] and [LEGACY-INTRODUCE1] below. Exactly one of the "enc-key ntor" and "enc-key legacy" elements must be present for each introduction point. [TODO: I'd like to have a cross-certification here too.] Other encryption and authentication key formats are allowed; clients should ignore ones they do not recognize. 3. The introduction protocol The introduction protocol proceeds in three steps. First, a hidden service host builds an anonymous circuit to a Tor node and registers that circuit as an introduction point. [Between these steps, the hidden service publishes its introduction points and associated keys, and the client fetches them as described in section [HSDIR] above.] Second, a client builds an anonymous circuit to the introduction point, and sends an introduction request. Third, the introduction point relays the introduction request along the introduction circuit to the hidden service host, and acknowledges the introduction request to the client. 3.1. Registering an introduction point [REG_INTRO_POINT] 3.1.1. Extensible ESTABLISH_INTRO protocol. [EST_INTRO] When a hidden service is establishing a new introduction point, it sends a ESTABLISH_INTRO cell with the following contents: AUTH_KEY_TYPE [1 byte] AUTH_KEY_LEN [1 byte] AUTH_KEY [AUTH_KEY_LEN bytes] Any number of times: EXT_FIELD_TYPE [1 byte] EXT_FIELD_LEN [1 byte] EXT_FIELD [EXTRA_FIELD_LEN bytes] ZERO [1 byte] HANDSHAKE_AUTH [MAC_LEN bytes] SIGLEN [1 byte] SIG [SIGLEN bytes] The AUTH_KEY_TYPE field indicates the type of the introduction point authentication key and the type of the MAC to use in for HANDSHAKE_AUTH. Recognized types are: [00, 01] -- Reserved for legacy introduction cells; see [LEGACY_EST_INTRO below] [02] -- Ed25519; HMAC-SHA256. [FF] -- Reserved for maintenance messages on existing circuits; see MAINT_INTRO below. [TODO: Should this just be a new relay cell type? Matthew and George think so.] The AUTH_KEY_LEN field determines the length of the AUTH_KEY field. The AUTH_KEY field contains the public introduction point authentication key. The EXT_FIELD_TYPE, EXT_FIELD_LEN, EXT_FIELD entries are reserved for future extensions to the introduction protocol. Extensions with unrecognized EXT_FIELD_TYPE values must be ignored. The ZERO field contains the byte zero; it marks the end of the extension fields. The HANDSHAKE_AUTH field contains the MAC of all earlier fields in the cell using as its key the shared per-circuit material ("KH") generated during the circuit extension protocol; see tor-spec.txt section 5.2, "Setting circuit keys". It prevents replays of ESTABLISH_INTRO cells. SIGLEN is the length of the signature. SIG is a signature, using AUTH_KEY, of all contents of the cell, up to but not including SIG. These contents are prefixed with the string "Tor establish-intro cell v1". Upon receiving an ESTABLISH_INTRO cell, a Tor node first decodes the key and the signature, and checks the signature. The node must reject the ESTABLISH_INTRO cell and destroy the circuit in these cases: * If the key type is unrecognized * If the key is ill-formatted * If the signature is incorrect * If the HANDSHAKE_AUTH value is incorrect * If the circuit is already a rendezvous circuit. * If the circuit is already an introduction circuit. [TODO: some scalability designs fail there.] * If the key is already in use by another circuit. Otherwise, the node must associate the key with the circuit, for use later in INTRODUCE1 cells. [TODO: The above will work fine with what we do today, but it will do quite badly if we ever freak out and want to go back to RSA2048 or bigger. Do we care?] 3.1.2. Registering an introduction point on a legacy Tor node [LEGACY_EST_INTRO] Tor nodes should also support an older version of the ESTABLISH_INTRO cell, first documented in rend-spec.txt. New hidden service hosts must use this format when establishing introduction points at older Tor nodes that do not support the format above in [EST_INTRO]. In this older protocol, an ESTABLISH_INTRO cell contains: KEY_LENGTH [2 bytes] KEY [KEY_LENGTH bytes] HANDSHAKE_AUTH [20 bytes] SIG [variable, up to end of relay payload] The KEY_LENGTH variable determines the length of the KEY field. The KEY field is a ASN1-encoded RSA public key. The HANDSHAKE_AUTH field contains the SHA1 digest of (KH | "INTRODUCE"). The SIG field contains an RSA signature, using PKCS1 padding, of all earlier fields. Note that since the relay payload itself may be no more than 498 bytes long, the KEY_LENGTH field can never have a first byte other than [00] or [01]. These values are used to distinguish legacy ESTABLISH_INTRO cells from newer ones. Older versions of Tor always use a 1024-bit RSA key for these introduction authentication keys. Newer hidden services MAY use RSA keys up 1904 bits. Any more than that will not fit in a RELAY cell payload. 3.1.3. Managing introduction circuits [MAINT_INTRO] If the first byte of an ESTABLISH_INTRO cell is [FF], the cell's body contains an administrative command for the circuit. The format of such a command is: Any number of times: SUBCOMMAND_TYPE [2 bytes] SUBCOMMAND_LEN [2 bytes] SUBCOMMAND [COMMAND_LEN bytes] Recognized SUBCOMMAND_TYPE values are: [00 01] -- update encryption keys [TODO: Matthew says, "This can be used to fork an intro point to balance traffic over multiple hidden service servers while maintaining the criteria for a valid ESTABLISH_INTRO cell. -MF". Investigate.] Unrecognized SUBCOMMAND_TYPE values should be ignored. 3.1.3.1. Updating encryption keys (subcommand 0001) [UPDATE-KEYS-SUBCMD] Hidden service hosts send this subcommand to set their initial encryption keys or update the configured public encryption keys associated with this circuit. This message must be sent after establishing an introduction point, before the circuit can be advertised. These keys are given in the form: NUMKEYS [1 byte] NUMKEYS times: KEYTYPE [1 byte] KEYLEN [1 byte] KEY [KEYLEN bytes] COUNTER [4 bytes] SIGLEN [1 byte] SIGNATURE [SIGLEN bytes.] The KEYTYPE value [01] is for Curve25519 keys. The COUNTER field is a monotonically increasing value across a given introduction point authentication key. The SIGNATURE must be generated with the introduction point authentication key, and must cover the entire subcommand body, prefixed with the string "Tor hidden service introduction encryption keys v1". [TODO: Nothing is done here to prove ownership of the encryption keys. Does that matter?] [TODO: The point here is to allow encryption keys to change while maintaining an introduction point and not forcing a client to download a new descriptor. I'm not sure if that's worth it. It makes clients who have seen a key before distinguishable from ones who have not.] [Matthew says: "Repeat-client over long periods of time will always be distinguishable. It may be better to simply expire intro points than try to preserve forward-secrecy, though". Must find out what he meant.] Setting the encryption keys for a given circuit replaces the previous keys for that circuit. Clients who attempt to connect using the old key receive an INTRO_ACK cell with error code [00 02] as described in section [INTRO_ACK] below. 3.1.4. Acknowledging establishment of introduction point [INTRO_ESTABLISHED] After setting up an introduction circuit, the introduction point reports its status back to the hidden service host with an empty INTRO_ESTABLISHED cell. [TODO: make this cell type extensible. It should be able to include data if that turns out to be needed.] 3.2. Sending an INTRODUCE1 cell to the introduction point. [SEND_INTRO1] In order to participate in the introduction protocol, a client must know the following: * An introduction point for a service. * The introduction authentication key for that introduction point. * The introduction encryption key for that introduction point. The client sends an INTRODUCE1 cell to the introduction point, containing an identifier for the service, an identifier for the encryption key that the client intends to use, and an opaque blob to be relayed to the hidden service host. In reply, the introduction point sends an INTRODUCE_ACK cell back to the client, either informing it that its request has been delivered, or that its request will not succeed. 3.2.1. INTRODUCE1 cell format [FMT_INTRO1] An INTRODUCE1 cell has the following contents: AUTH_KEYID [32 bytes] ENC_KEYID [8 bytes] Any number of times: EXT_FIELD_TYPE [1 byte] EXT_FIELD_LEN [1 byte] EXT_FIELD [EXTRA_FIELD_LEN bytes] ZERO [1 byte] ENCRYPTED [Up to end of relay payload] [TODO: Should we have a field to determine the type of ENCRYPTED, or should we instead assume that there is exactly one encryption key per encryption method? The latter is probably safer.] Upon receiving an INTRODUCE1 cell, the introduction point checks whether AUTH_KEYID and ENC_KEYID match a configured introduction point authentication key and introduction point encryption key. If they do, the cell is relayed; if not, it is not. The AUTH_KEYID for an Ed25519 public key is the public key itself. The ENC_KEYID for a Curve25519 public key is the first 8 bytes of the public key. (This key ID is safe to truncate, since all the keys are generated by the hidden service host, and the ID is only valid relative to a single AUTH_KEYID.) The ENCRYPTED field is as described in 3.3 below. To relay an INTRODUCE1 cell, the introduction point sends an INTRODUCE2 cell with exactly the same contents. 3.2.2. INTRODUCE_ACK cell format. [INTRO_ACK] An INTRODUCE_ACK cell has the following fields: STATUS [2 bytes] Any number of times: EXT_FIELD_TYPE [1 byte] EXT_FIELD_LEN [1 byte] EXT_FIELD [EXTRA_FIELD_LEN bytes] Recognized status values are: [00 00] -- Success: cell relayed to hidden service host. [00 01] -- Failure: service ID not recognzied [00 02] -- Failure: key ID not recognized [00 03] -- Bad message format Recognized extension field types: [00 01] -- signed set of encryption keys The extension field type 0001 is a signed set of encryption keys; its body matches the body of the key update command in [UPDATE-KEYS-CMD]. Whenever sending status [00 02], the introduction point MUST send this extension field. 3.2.3. Legacy formats [LEGACY-INTRODUCE1] When the ESTABLISH_INTRO cell format of [LEGACY_EST_INTRO] is used, INTRODUCE1 cells are of the form: AUTH_KEYID_HASH [20 bytes] ENC_KEYID [8 bytes] Any number of times: EXT_FIELD_TYPE [1 byte] EXT_FIELD_LEN [1 byte] EXT_FIELD [EXTRA_FIELD_LEN bytes] ZERO [1 byte] ENCRYPTED [Up to end of relay payload] Here, AUTH_KEYID_HASH is the hash of the introduction point authentication key used to establish the introduction. Because of limitations in older versions of Tor, the relay payload size for these INTRODUCE1 cells must always be at least 246 bytes, or they will be rejected as invalid. 3.3. Processing an INTRODUCE2 cell at the hidden service. [PROCESS_INTRO2] Upon receiving an INTRODUCE2 cell, the hidden service host checks whether the AUTH_KEYID/AUTH_KEYID_HASH field and the ENC_KEYID fields are as expected, and match the configured authentication and encryption key(s) on that circuit. The service host then checks whether it has received a cell with these contents before. If it has, it silently drops it as a replay. (It must maintain a replay cache for as long as it accepts cells with the same encryption key.) If the cell is not a replay, it decrypts the ENCRYPTED field, establishes a shared key with the client, and authenticates the whole contents of the cell as having been unmodified since they left the client. There may be multiple ways of decrypting the ENCRYTPED field, depending on the chosen type of the encryption key. Requirements for an introduction handshake protocol are described in [INTRO-HANDSHAKE-REQS]. We specify one below in section [NTOR-WITH-EXTRA-DATA]. The decrypted plaintext must have the form: REND_TOKEN [20 bytes] Any number of times: EXT_FIELD_TYPE [1 byte] EXT_FIELD_LEN [1 byte] EXT_FIELD [EXTRA_FIELD_LEN bytes] ZERO [1 byte] ONION_KEY_TYPE [2 bytes] ONION_KEY [depends on ONION_KEY_TYPE] NSPEC (Number of link specifiers) [1 byte] NSPEC times: LSTYPE (Link specifier type) [1 byte] LSLEN (Link specifier length) [1 byte] LSPEC (Link specifier) [LSLEN bytes] PAD (optional padding) [up to end of plaintext] Upon processing this plaintext, the hidden service makes sure that any required authentication is present in the extension fields, and then extends a rendezvous circuit to the node described in the LSPEC fields, using the ONION_KEY to complete the extension. As mentioned in [BUILDING-BLOCKS], the "TLS-over-TCP, IPv4" and "Legacy node identity" specifiers must be present. The hidden service SHOULD NOT reject any LSTYPE fields which it doesn't recognize; instead, it should use them verbatim in its EXTEND request to the rendezvous point. The ONION_KEY_TYPE field is one of: [01] TAP-RSA-1024: ONION_KEY is 128 bytes long. [02] NTOR: ONION_KEY is 32 bytes long. The ONION_KEY field describes the onion key that must be used when extending to the rendezvous point. It must be of a type listed as supported in the hidden service descriptor. Upon receiving a well-formed INTRODUCE2 cell, the hidden service host will have: * The information needed to connect to the client's chosen rendezvous point. * The second half of a handshake to authenticate and establish a shared key with the hidden service client. * A set of shared keys to use for end-to-end encryption. 3.3.1. Introduction handshake encryption requirements [INTRO-HANDSHAKE-REQS] When decoding the encrypted information in an INTRODUCE2 cell, a hidden service host must be able to: * Decrypt additional information included in the INTRODUCE2 cell, to include the rendezvous token and the information needed to extend to the rendezvous point. * Establish a set of shared keys for use with the client. * Authenticate that the cell has not been modified since the client generated it. Note that the old TAP-derived protocol of the previous hidden service design achieved the first two requirements, but not the third. 3.3.2. Example encryption handshake: ntor with extra data [NTOR-WITH-EXTRA-DATA] This is a variant of the ntor handshake (see tor-spec.txt, section 5.1.4; see proposal 216; and see "Anonymity and one-way authentication in key-exchange protocols" by Goldberg, Stebila, and Ustaoglu). It behaves the same as the ntor handshake, except that, in addition to negotiating forward secure keys, it also provides a means for encrypting non-forward-secure data to the server (in this case, to the hidden service host) as part of the handshake. Notation here is as in section 5.1.4 of tor-spec.txt, which defines the ntor handshake. The PROTOID for this variant is "hidden-service-ntor-curve25519-sha256-1". Define the tweak value t_hsenc, and the tag value m_hsexpand as: t_hsenc = PROTOID | ":hs_key_extract" m_hsexpand = PROTOID | ":hs_key_expand" To make an INTRODUCE cell, the client must know a public encryption key B for the hidden service on this introduction circuit. The client generates a single-use keypair: x,X = KEYGEN() and computes: secret_hs_input = EXP(B,x) | AUTH_KEYID | X | B | PROTOID info = m_hsexpand | subcredential hs_keys = HKDF(secret_hs_input, t_hsenc, info, S_KEY_LEN+MAC_LEN) ENC_KEY = hs_keys[0:S_KEY_LEN] MAC_KEY = hs_keys[S_KEY_LEN:S_KEY_LEN+MAC_KEY_LEN] and sends, as the ENCRYPTED part of the INTRODUCE1 cell: CLIENT_PK [G_LENGTH bytes] ENCRYPTED_DATA [Padded to length of plaintext] MAC [MAC_LEN bytes] Substituting those fields into the INTRODUCE1 cell body format described in [FMT_INTRO1] above, we have AUTH_KEYID [32 bytes] ENC_KEYID [8 bytes] Any number of times: EXT_FIELD_TYPE [1 byte] EXT_FIELD_LEN [1 byte] EXT_FIELD [EXTRA_FIELD_LEN bytes] ZERO [1 byte] ENCRYPTED: CLIENT_PK [G_LENGTH bytes] ENCRYPTED_DATA [Padded to length of plaintext] MAC [MAC_LEN bytes] (This format is as documented in [FMT_INTRO1] above, except that here we describe how to build the ENCRYPTED portion. If the introduction point is running an older Tor that does not support this protocol, the first field is replaced by a 20-byte AUTH_KEYID_HASH field as described in [LEGACY-INTRODUCE1].) Here, the encryption key plays the role of B in the regular ntor handshake, and the AUTH_KEYID field plays the role of the node ID. The CLIENT_PK field is the public key X. The ENCRYPTED_DATA field is the message plaintext, encrypted with the symmetric key ENC_KEY. The MAC field is a MAC of all of the cell from the AUTH_KEYID through the end of ENCRYPTED_DATA, using the MAC_KEY value as its key. To process this format, the hidden service checks PK_VALID(CLIENT_PK) as necessary, and then computes ENC_KEY and MAC_KEY as the client did above, except using EXP(CLIENT_PK,b) in the calculation of secret_hs_input. The service host then checks whether the MAC is correct. If it is invalid, it drops the cell. Otherwise, it computes the plaintext by decrypting ENCRYPTED_DATA. The hidden service host now completes the service side of the extended ntor handshake, as described in tor-spec.txt section 5.1.4, with the modified PROTOID as given above. To be explicit, the hidden service host generates a keypair of y,Y = KEYGEN(), and uses its introduction point encryption key 'b' to computes: xb = EXP(X,b) secret_hs_input = xb | AUTH_KEYID | X | B | PROTOID info = m_hsexpand | subcredential hs_keys = HKDF(secret_hs_input, t_hsenc, info, S_KEY_LEN+MAC_LEN) HS_DEC_KEY = hs_keys[0:S_KEY_LEN] HS_MAC_KEY = hs_keys[S_KEY_LEN:S_KEY_LEN+MAC_KEY_LEN] (The above are used to check the MAC and then decrypt the encrypted data.) ntor_secret_input = EXP(X,y) | xb | ID | B | X | Y | PROTOID NTOR_KEY_SEED = H(secret_input, t_key) verify = H(secret_input, t_verify) auth_input = verify | ID | B | Y | X | PROTOID | "Server" (The above are used to finish the ntor handshake.) The server's handshake reply is: SERVER_PK Y [G_LENGTH bytes] AUTH H(auth_input, t_mac) [H_LENGTH bytes] These faileds can be send to the client in a RENDEZVOUS1 cell. (See [JOIN_REND] below.) The hidden service host now also knows the keys generated by the handshake, which it will use to encrypt and authenticate data end-to-end between the client and the server. These keys are as computed in tor-spec.txt section 5.1.4. 3.4. Authentication during the introduction phase. [INTRO-AUTH] Hidden services may restrict access only to authorized users. One mechanism to do so is the credential mechanism, where only users who know the credential for a hidden service may connect at all. For more fine-grained conntrol, a hidden service can be configured with password-based or public-key-based authentication. 3.4.1. Password-based authentication. To authenticate with a password, the user must include an extension field in the encrypted part of the INTRODUCE cell with an EXT_FIELD_TYPE type of [01] and the contents: Username [00] Password. The username may not include any [00] bytes. The password may. On the server side, the password MUST be stored hashed and salted, ideally with scrypt or something better. 3.4.2. Ed25519-based authentication. To authenticate with an Ed25519 private key, the user must include an extension field in the encrypted part of the INTRODUCE cell with an EXT_FIELD_TYPE type of [02] and the contents: Nonce [16 bytes] Pubkey [32 bytes] Signature [64 bytes] Nonce is a random value. Pubkey is the public key that will be used to authenticate. [TODO: should this be an identifier for the public key instead?] Signature is the signature, using Ed25519, of: "Hidserv-userauth-ed25519" Nonce (same as above) Pubkey (same as above) AUTH_KEYID (As in the INTRODUCE1 cell) ENC_KEYID (As in the INTRODUCE1 cell) The hidden service host checks this by seeing whether it recognizes and would accept a signature from the provided public key. If it would, then it checks whether the signature is correct. If it is, then the correct user has authenticated. Replay prevention on the whole cell is sufficient to prevent replays on the authentication. Users SHOULD NOT use the same public key with multiple hidden services. 4. The rendezvous protocol Before connecting to a hidden service, the client first builds a circuit to an arbitrarily chosen Tor node (known as the rendezvous point), and sends an ESTABLISH_RENDEZVOUS cell. The hidden service later connects to the same node and sends a RENDEZVOUS cell. Once this has occurred, the relay forwards the contents of the RENDEZVOUS cell to the client, and joins the two circuits together. 4.1. Establishing a rendezvous point [EST_REND_POINT] The client sends the rendezvous point a RELAY_COMMAND_ESTABLISH_RENDEZVOUS cell containing a 20-byte value. RENDEZVOUS_COOKIE [20 bytes] Rendezvous points MUST ignore any extra bytes in an ESTABLISH_RENDEZVOUS message. (Older versions of Tor did not.) The rendezvous cookie is an arbitrary 20-byte value, chosen randomly by the client. The client SHOULD choose a new rendezvous cookie for each new connection attempt. If the rendezvous cookie is already in use on an existing circuit, the rendezvous point should reject it and destroy the circuit. Upon receiving a ESTABLISH_RENDEZVOUS cell, the rendezvous point associates the cookie with the circuit on which it was sent. It replies to the client with an empty RENDEZVOUS_ESTABLISHED cell to indicate success. [TODO: make this extensible] The client MUST NOT use the circuit which sent the cell for any purpose other than rendezvous with the given location-hidden service. The client should establish a rendezvous point BEFORE trying to connect to a hidden service. 4.2. Joining to a rendezvous point [JOIN_REND] To complete a rendezvous, the hidden service host builds a circuit to the rendezvous point and sends a RENDEZVOUS1 cell containing: RENDEZVOUS_COOKIE [20 bytes] HANDSHAKE_INFO [variable; depends on handshake type used.] If the cookie matches the rendezvous cookie set on any not-yet-connected circuit on the rendezvous point, the rendezvous point connects the two circuits, and sends a RENDEZVOUS2 cell to the client containing the contents of the RENDEZVOUS1 cell. Upon receiving the RENDEZVOUS2 cell, the client verifies that the HANDSHAKE_INFO correctly completes a handshake, and uses the handshake output to derive shared keys for use on the circuit. [TODO: Should we encrypt HANDSHAKE_INFO as we did INTRODUCE2 contents? It's not necessary, but it could be wise. Similarly, we should make it extensible.] 4.3. Using legacy hosts as rendezvous points The behavior of ESTABLISH_RENDEZVOUS is unchanged from older versions of this protocol, except that relays should now ignore unexpected bytes at the end. Old versions of Tor required that RENDEZVOUS cell payloads be exactly 168 bytes long. All shorter rendezvous payloads should be padded to this length with [00] bytes. 5. Encrypting data between client and host A successfully completed handshake, as embedded in the INTRODUCE/RENDEZVOUS cells, gives the client and hidden service host a shared set of keys Kf, Kb, Df, Db, which they use for sending end-to-end traffic encryption and authentication as in the regular Tor relay encryption protocol, applying encryption with these keys before other encryption, and decrypting with these keys before other encryption. The client encrypts with Kf and decrypts with Kb; the service host does the opposite. 6. Open Questions: Scaling hidden services is hard. There are on-going discussions that you might be able to help with. See [SCALING-REFS]. How can we improve the HSDir unpredictability design proposed in [SHAREDRANDOM]? See [SHAREDRANDOM-REFS] for discussion. How can hidden service addresses become memorable while retaining their self-authenticating and decentralized nature? See [HUMANE-HSADDRESSES-REFS] for some proposals; many more are possible. Hidden Services are pretty slow. Both because of the lengthy setup procedure and because the final circuit has 6 hops. How can we make the Hidden Service protocol faster? See [PERFORMANCE-REFS] for some suggestions. References: [KEYBLIND-REFS]: https://trac.torproject.org/projects/tor/ticket/8106 https://lists.torproject.org/pipermail/tor-dev/2012-September/004026.html [SHAREDRANDOM-REFS]: https://trac.torproject.org/projects/tor/ticket/8244 https://lists.torproject.org/pipermail/tor-dev/2013-November/005847.html https://lists.torproject.org/pipermail/tor-talk/2013-November/031230.html [SCALING-REFS]: https://lists.torproject.org/pipermail/tor-dev/2013-October/005556.html [HUMANE-HSADDRESSES-REFS]: https://gitweb.torproject.org/torspec.git/blob/HEAD:/proposals/ideas/xxx-on… http://archives.seul.org/or/dev/Dec-2011/msg00034.html [PERFORMANCE-REFS]: "Improving Efficiency and Simplicity of Tor circuit establishment and hidden services" by Overlier, L., and P. Syverson [TODO: Need more here! Do we have any? :( ] [ATTACK-REFS]: "Trawling for Tor Hidden Services: Detection, Measurement, Deanonymization" by Alex Biryukov, Ivan Pustogarov, Ralf-Philipp Weinmann "Locating Hidden Servers" by Lasse Øverlier and Paul Syverson [ED25519-REFS]: "High-speed high-security signatures" by Daniel J. Bernstein, Niels Duif, Tanja Lange, Peter Schwabe, and Bo-Yin Yang. http://cr.yp.to/papers.html#ed25519 Appendix A. Signature scheme with key blinding [KEYBLIND] As described in [IMD:DIST] and [SUBCRED] above, we require a "key blinding" system that works (roughly) as follows: There is a master keypair (sk, pk). Given the keypair and a nonce n, there is a derivation function that gives a new blinded keypair (sk_n, pk_n). This keypair can be used for signing. Given only the public key and the nonce, there is a function that gives pk_n. Without knowing pk, it is not possible to derive pk_n; without knowing sk, it is not possible to derive sk_n. It's possible to check that a signature make with sk_n while knowing only pk_n. Someone who sees a large number of blinded public keys and signatures made using those public keys can't tell which signatures and which blinded keys were derived from the same master keypair. You can't forge signatures. [TODO: Insert a more rigorous definition and better references.] We propose the following scheme for key blinding, based on Ed25519. (This is an ECC group, so remember that scalar multiplication is the trapdoor function, and it's defined in terms of iterated point addition. See the Ed25519 paper [Reference ED25519-REFS] for a fairly clear writeup.) Let the basepoint be written as B. Assume B has prime order l, so lB=0. Let a master keypair be written as (a,A), where a is the private key and A is the public key (A=aB). To derive the key for a nonce N and an optional secret s, compute the blinding factor h as H(A | s, B, N), and let: private key for the period: a' = h a public key for the period: A' = h' A = (ha)B Generating a signature of M: given a deterministic random-looking r (see EdDSA paper), take R=rB, S=r+hash(R,A',M)ah mod l. Send signature (R,S) and public key A'. Verifying the signature: Check whether SB = R+hash(R,A',M)A'. (If the signature is valid, SB = (r + hash(R,A',M)ah)B = rB + (hash(R,A',M)ah)B = R + hash(R,A',M)A' ) See [KEYBLIND-REFS] for an extensive discussion on this scheme and possible alternatives. I've transcribed this from a description by Tanja Lange at the end of the thread. [TODO: We'll want a proof for this.] (To use this with Tor, set N = INT_8(period-number) | INT_8(Start of period in seconds since epoch).) Appendix B. Selecting nodes [PICKNODES] Picking introduction points Picking rendezvous points Building paths Reusing circuits (TODO: This needs a writeup) Appendix C. Recommendations for searching for vanity .onions [VANITY] EDITORIAL NOTE: The author thinks that it's silly to brute-force the keyspace for a key that, when base-32 encoded, spells out the name of your website. It also feels a bit dangerous to me. If you train your users to connect to llamanymityx4fi3l6x2gyzmtmgxjyqyorj9qsb5r543izcwymle.onion I worry that you're making it easier for somebody to trick them into connecting to llamanymityb4sqi0ta0tsw6uovyhwlezkcrmczeuzdvfauuemle.onion Nevertheless, people are probably going to try to do this, so here's a decent algorithm to use. To search for a public key with some criterion X: Generate a random (sk,pk) pair. While pk does not satisfy X: Add the number 1 to sk Add the scalar B to pk Return sk, pk. This algorithm is safe [source: djb, personal communication] [TODO: Make sure I understood correctly!] so long as only the final (sk,pk) pair is used, and all previous values are discarded. To parallelize this algorithm, start with an independent (sk,pk) pair generated for each independent thread, and let each search proceed independently. Appendix D. Numeric values reserved in this document [TODO: collect all the lists of commands and values mentioned above]

8 22

Call for a big fast bridge (to be the meek backend)
by David Fifield 21 Mar '15

21 Mar '15

The meek pluggable transport is currently running on the bridge I run, which also happens to be the backend bridge for flash proxy. I'd like to move it to a fast relay run by an experienced operator. I want to do this both to diffuse trust, so that I don't run all the infrastructure, and because my bridge is not especially fast and I'm not especially adept at performance tuning. All you will need to do is run the meek-server program, add some lines to your torrc, and update the software when I ask you to. The more CPU, memory, and bandwidth you have, the better, though at this point usage is low enough that you won't even notice it if you are already running a fast relay. I think it will help if your bridge is located in the U.S., because that reduces latency from Google App Engine. The meek-server plugin is basically just a little web server: https://gitweb.torproject.org/pluggable-transports/meek.git/tree/HEAD:/meek… Since meek works differently than obfs3, for example, it doesn't help us to have hundreds of medium-fast bridges. We need one (or maybe two or three) big fat fast relays, because all the traffic that is bounced through App Engine or Amazon will be pointed at it. My PGP key is at https://www.bamsoftware.com/david/david.asc if you want to talk about it. David Fifield

6 11

Hidden service policies
by Mike Hearn 30 Dec '14

30 Dec '14

Hello, As we know, hidden services can be useful for all kinds of legitimate things (Pond's usage is particularly interesting), however they do also sometimes get used by botnets and other problematic things. Tor provides exit policies to let exit relay operators restrict traffic they consider to be unwanted or abusive. In this way a kind of international group consensus emerges about what is and is not acceptable usage of Tor. For instance, SMTP out is widely restricted. Has there been any discussion of implementing similar controls for hidden services, where relays would refuse to act as introduction points for hidden services that match certain criteria e.g. have a particular key, or whose key appears in a list downloaded occasionally via Tor itself. In this way relay operators could avoid their resources being used for establishing communication with botnet CnC servers. Obviously such a scheme would require a protocol and client upgrade to avoid nodes building circuits to relays that then refuse to introduce. The downside is additional complexity. The upside is potentially recruiting new relay operators.

6 11

Defending against guard discovery attacks by pinning middle nodes
by George Kadianakis 13 Nov '14

13 Nov '14

Hey Nick, this mail is about the schemes we were discussing during the dev meeting on how to protect HSes against guard discovery attacks (#9001). I think we have some ideas on how to offer better protection against such attacks, mainly by keeping our middle nodes more static than we do currently. For example, we could keep our middle nodes for 3-4 days instead of choosing new ones for every circuit. As Roger has suggested, maybe we don't even need to write the static middle nodes on the state file, just use new ones if Tor has restarted. Keeping middle nodes around for longer will make those attacks much slower (it restricts them to one attack attempt every 3-4 days), but are there any serious negative implications? For example, if you were unlucky and you picked an evil middle node, and you keep it for 3-4 days, that middle node will always see your traffic coming through your guard (assuming a single guard per client). If we assume you use a non-popular guard node (with only a few clients using it), the middle guard might be able to think "Ah, the circuit that comes from that guard node is always user X" making your circuits a bit linkable from the PoV of your middle node. What other attacks should we be wary about? Maybe partitioning attacks based on client behavior? And how should we move this forward if we decide it's worth it? Should we start writing a Tor proposal? Thanks!

8 26

Request to expunge project 'torsocks' on Google Code
by Virgil Griffith 26 Oct '14

26 Oct '14

Sitting with David Goulet, it's problematic that when I search for "torsocks" I get the old Google Code page instead of the up-to-date page on http://gitweb.torproject.org This is problematic because people report bugs to the Google Code repository instead of the torproject repository. Jake/Robert: You are the project owners on Google Code. Could you please remove the project? Much love, -Virgil

6 8

Reminder: weekly Tor development meeting (Wed), and weekly patch workshop (Tue)
by Nick Mathewson 21 Oct '14

21 Oct '14

Hi, all! You probably know that we do our regular meetings for tor (the program) development on Wednesdays at 1330 UTC on the #tor-dev IRC channel on irc.oftc.net. But did you know everybody who's interested in doing anything with tor is welcome to attend? And did you know that anybody who's interested is welcome to lurk? Also, we've started doing a weekly patch workshop where we all look at each other's patches and try to help get them done, try to help code get better, and so forth. That's on Tuesdays at 1700 UTC on the same IRC channel! (Thanks for rl1987 for suggesting this.) cheers, -- Nick

1 1

Guardiness: Yet another external dirauth script
by George Kadianakis 09 Oct '14

09 Oct '14

==Guardiness: Yet another external dirauth script== ====Introduction==== One well-known problem with Tor relays, is that Guards will suffer a big loss of traffic as soon as they get the Guard flag. This happens because clients pick guards every 2-3 months, so young guards will not get picked by old clients and mainly attract new clients. This is documented in 'phase three' of Roger's blog post: https://blog.torproject.org/blog/lifecycle-of-a-new-relay The problem gets even worse if we extend the guard lifetime to 8-9 months. The plan to solve this problem is to make client load balancing a bit smarter by priotizing guards that suffer this traffic loss as middle relays. The reason I'm sending this email is because this feature is by far the trickiest part of prop236 (guard node security) and I wanted to inform all dirauths of our plan and ask for feedback on the deployment procedure. ====How guardiness works==== Authorities calculate for each relay how many consensuses it has been a guard for the past 2-3 months, and then they note that fraction down in the consensus. Then clients parse the consensus and if they see a guard that has been a guard for 55% of the past consensuses, they will consider that relay as 55% guard and 45% non-guard (that's 100% - 55%). You can find more information at: https://gitweb.torproject.org/torspec.git/blob/HEAD:/proposals/236-single-g… The idea was that the guardiness script will be an external script that is run by Tor in a similar fashion to the bandwidth auth scripts. We chose that because we could write the script in a high-level language and because it could be modular and we could change the algorithm in the future if we wanted. Unfortunately, it seems that external scripts in dirauths are a PITA to maintain as can be seen by the lack of bwauth operators. ====How the guardiness script works==== The guardiness script, is supposed to parse 2-3 months worth of consensuses (but should also be to do the same for 9 months worth of consensuses) and calculate the guard fraction of each guard, save it to a file, and have the dirauth read it to update its routerstatuses. One problem I encountered from early on, is that stem takes about 30mins to parse 3 months of consesuses (~2000 consensuses). Since this script should ideally be run every hour before each authority votes, such long parsing time is unacceptable. I mentioned this problem at https://trac.torproject.org/projects/tor/ticket/9321#comment:19 and stated a few possible solutions. I received some feedback from Nick, and the solution I decided to take in the end is to have another script that is called first and summarizes consensuses to summary files. Summary files are then saved to disk, and parsed by the guardiness script to produce an output file that is read by dirauths. Summary files are designed to be quick to parse (even with Python) and contain all the necessary information for guardiness. For example, parsing 2000 summary files in my laptop takes about 10 seconds. FWIW, the guardiness scripts are ready for review and can be found here: https://gitweb.torproject.org/user/asn/hax.git/shortlog/refs/heads/guardine… ====How the guardiness script will be deployed==== The idea is that dirauths will add another script to their crontab that is called every hour (before or after the bwauth scripts). The script first calls the summarizer script, which goes to the consensus/ directory and summarizes all consensuses it finds and puts them in the summary/ directory. The summarizer script then deletes all the consensuses that got summarized. Then the script calls the the guardiness script, which goes to the summary/ directory, parses all summary files it finds, and outputs a guardiness output file that gets parsed by the dirauth prior to voting. That should be all. Easy, eh? :) Now I will start a FAQ section where I state my doubts and fears. ====FAQ==== - Q: Where do dirauths find all those old consensuses? There are various ways for dirauths to populate their consensus/ directory. They could fetch consensuses from metrics, or they could add a cron job that copies cached-consensus to a directory every hour. However, I think the cleanest solution is to use Daniel Martí's upcoming consensus diff changes. Daniel will add a torrc option that allows Tor to save consensuses to a directory. My idea was to get dirauths to use Daniel's code to populate their consensus/ directory for two or three months. And then, after two or three months enable the guardiness scripts. To make sure that this is indeed the best approach, I need to learn from Nick when he plans to merge Daniel's code to Tor. - Q: How does guardiness look like in the consensus? Here is how a guard with guardiness (GuardFraction) 10% looks like in the consensus: r test006r HyS1DRHzEojbQVPZ1B3zAHc/HY0 9St4yWfV4huz5V86mt24HL3Yi2I 2014-09-06 13:44:28 127.0.0.1 5006 7006 s Exit Fast Guard HSDir Running Stable V2Dir Valid v Tor 0.2.6.0-alpha-dev w Bandwidth=111 Unmeasured=1 GuardFraction=10 - Q: What are you afraid of? I'm mainly afraid of misconfiguration problems. This guardiness system is a bit complex and I'm not expecting dirauths to learn how to use it and debug it, so it should work easily and well... Here are some specific issues: -- File management For example, I'm afraid of the file management mess that summary files cause. We need to make sure that we don't leave old consensuses/summary files rot in the filesystem. Or that we don't summarize the same consensuses over and over again. To do that, I added some optional cleanup switches to both scripts: Specifically, the summarizer script can delete consensus files that already got summarized and can also delete consensus files older than 3 months (or N months). Similarly, the guardiness.py script can delete summary files older than 3 months (or N months). The idea is that every time the cron job triggers, both scripts will auto-delete the oldest summary/consensus file, keeping in disk only the useful files. -- Incomplete consensus data set I'm afraid that a directory authority might not have a properly populated consensus directory and hence advertise wrong guard fractions. For example, maybe it only has 10 consensuses in its consensus directory instead of 1900 consensuses. Since the authorities only state the guardiness percentage in the consensus, it's not possible to learn how many consensuses were in their dataset. Maybe we need to add a "guardiness-consensus-parsed" in their votes, to easier debug such issues? Also, 3 months worth of consensuses is 2160 consensuses. Because dirauths sometimes misbehave, it's certain that not all 2160 consensuses will have been issued and that's normal. But how do we understand if dirauths have a sufficiently good consensus data set? Is 2000 out of 2160 consensuses an OK data set? What about 1000 out of 2160 consensuses? Furthermore, we need to make sure that dirauths don't consider old consensuses in their GuardFraction calculations. To achieve this, both scripts have a mandatory switch that allows operators to specify the maximum consensus age that is acceptable. So for example, if you call the summarizer script with 3 months of consensus age, it will not parse consensuses older than 3 months. Furthermore, there is a CLI switch that allows the scripts to delete expired consensuses. - Q: Why do you slow stem instead of parsing consensuses with Python on your own? This is another part where I might have taken the wrong design decision, but I decided to not get into the consensus parsing business and just rely on stem. This is also because I was hoping to use stem to verify consensus signatures. However, now that we might use Daniel's patch to populate our consensus database, maybe we don't need to treat consensuses as untrusted anymore. If you think that I should try to parse the consensuses on my own, please tell me and I will give it a try. Maybe it will be fast. Definitely not as fast as summary files, but maybe we can parse 3 months worth of consesuses in 15 to 40 seconds. - Q: Why do you mess with multiple summary files and you don't just have *one* summary file? Because of the rolling nature of guardiness (we always want to consider the past 3 months), every hour we need to _discard_ the oldest observations (the consensus from 3 months ago) and start considering the newest consensus. Because we need to discard that oldest consensus, it's hard to keep information about each consensus in a single summary file. And that's why I chose to have a summary file for each consensus. Maybe it's the wrong decision though... - Q: What's up with the name "guardiness"? It's quite terrible I know but it was the name that we used from quite early on about this project. I think before finalizing this task I'm going to rename everything to 'GuardFraction' since it's more self-explanatory. I'm also considering names like "GuardTrafficLoadBalancer" etc.

3 8

Scaling tor for a global population
by Andrew Lewman 03 Oct '14

03 Oct '14

I had a conversation with a vendor yesterday. They are interested in including Tor as their "private browsing mode" and basically shipping a re-branded tor browser which lets people toggle the connectivity to the Tor network on and off. They very much like Tor Browser and would like to ship it to their customer base. Their product is 10-20% of the global market, this is of roughly 2.8 billion global Internet users. As Tor Browser is open source, they are already working on it. However ,their concern is scaling up to handling some percent of global users with "tor mode" enabled. They're willing to entertain offering their resources to help us solve the scalability challenges of handling hundreds of millions of users and relays on Tor. As this question keeps popping up by the business world looking at privacy as the next "must have" feature in their products, I'm trying to compile a list of tasks to solve to help us scale. The old 2008 three-year roadmap looks at performance, https://www.torproject.org/press/2008-12-19-roadmap-press-release.html.en I've been through the specs, https://gitweb.torproject.org/torspec.git/tree/HEAD:/proposals to see if there are proposals for scaling the network or directory authorities. I didn't see anything directly related. The last research paper I see directly addressing scalability is Torsk (http://www.freehaven.net/anonbib/bibtex.html#ccs09-torsk) or PIR-Tor (http://www.freehaven.net/anonbib/bibtex.html#usenix11-pirtor) Is there a better list available for someone new to Tor to read up on the scalability challenges? -- Andrew pgp 0x6B4D6475 https://www.torproject.org/ +1-781-948-1982

17 35

RFC: Trunnel -- a binary parser generator for Tor and more
by Nick Mathewson 02 Oct '14

02 Oct '14

Hi! Back when the Heartbleed vulnerability went public, I thought to myself: "If we keep generating writing C code to parse binary formats by hand, the same thing will happen to us sooner or later." So on and off, I've been working on a little tool to automatically generate binary encoding and parsing code based on C-like structure descriptions. It's a bit like nail[0], but more lightweight: it doesn't support backtracking, sub-byte alignment, or any of several other features that Tor doesn't actually want. You can see a pre-alpha version of the tool (called "Trunnel") in one of my git repositories [1]. The "doc" subdirectory has a simple manual [2]. To make sure that it would suit our needs, I wrote up trunnel specification files for the current Tor protocols, for SOCKS, and for the new protocols in proposals 220, 224, and 229. It seems to work okay. In case you want to see what the tool does without running it yourself, you can find the SOCKS+proposal229 specification here [3], and compare it with the generated header [4] and C implementation [5]. Trunnel is written in Python; my tests have 99% line coverage on the Python code. The C code which is generated by those tests itself has 99% line coverage from other unit tests, so I'm thinking that it's reasonably tested. (Branch coverage is lower, but still pretty high.) So, please check this out and have a look at it! My open questions are: * Is this a good idea? * Is it a good idea to use this in Tor? * Are there any tricky bugs left in the generated code? * What am I forgetting to think of? FWIW, my current plan is not to do a blanket replacement of our existing parse code with Trunnel (yet) but rather to use Trunnel to generate parsing code for new cell formats that are getting added in 0.2.6 and later. [0] https://github.com/jbangert/nail [1] https://gitweb.torproject.org/user/nickm/trunnel.git . [2] Also online at http://www.wangafu.net/~nickm/trunnel-manual.html [3] https://gitweb.torproject.org/user/nickm/trunnel.git/blob/HEAD:/examples/so… [4] http://www.wangafu.net/~nickm/volatile/socks5.h [5] http://www.wangafu.net/~nickm/volatile/socks5.c cheers, -- Nick Mathewson

1 1

Summary of meek's costs, August 2014
by David Fifield 01 Oct '14

01 Oct '14

The "meek" pluggable transport has started getting some users since my blog post two weeks ago (https://blog.torproject.org/blog/how-use-%E2%80%9Cmeek%E2%80%9D-pluggable-t…). See the attached file userstats-bridge-transport-meek-2014-08-01-2014-08-29.png. Today I logged in to the Google and Amazon accounts to see what the usage looks like. Here is a summary. In short, App Engine cost $1.02 this month, and Amazon cost $2.54. meek-appspot-2014-08-29.png shows the Google App Engine bytes/second for the month of August. App Engine costs $0.12 per GB, with 1 GB for free each day. It's finally at the point where it's costing money every day: All of January 2014: $ -- All of February 2014: $0.10 All of March 2014: $ -- All of April 2014: $0.83 All of May 2014: $0.75 All of June 2014: $0.73 All of July 2014: $0.59 August 2014 to date: $1.02 Total to date $4.02 Here are daily totals in the past two weeks: 2014-08-14 $ -- 2014-08-15 $ -- 2014-08-16 $ -- 2014-08-17 $ -- 2014-08-18 $0.09 2014-08-19 $0.00 2014-08-20 $0.04 2014-08-21 $0.07 2014-08-22 $0.09 2014-08-23 $0.07 2014-08-24 $0.03 2014-08-25 $0.09 2014-08-26 $0.19 2014-08-27 $0.07 2014-08-28 $0.28 meek-aws-2014-08-29.png shows the AWS megabytes per day (note that the y-axis is different from the other graph so they are not directly comparable). The charges break down quite differently: there's no cost for bandwidth so far, and all the charges are for the sheer number of HTTPS requests. The Amazon bill is broken down by region: Asia Pacific (Singapore) Region $0.0090 per 10,000 HTTP Requests 3 Requests $0.01 Asia Pacific (Sydney) Region $ -- EU (Ireland) Region $ -- South America (Sao Paulo) Region $ -- Asia Pacific (Tokyo) Region $0.0090 per 10,000 HTTP Requests 38 Requests $0.01 US East (Northern Virginia) Region $0.0100 per 10,000 Proxy HTTPS Requests (US) 2,513,557 Requests $2.51 $0.0075 per 10,000 HTTP Requests 3 Requests $0.01 Total to date $2.54 David Fifield

1 1

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

tor-dev September 2014