2.1.TLS Overview

TLS provides two endpoints with a way to establish a means of communication overan untrusted medium (for example, the Internet). TLS enables authentication ofpeers and provides confidentiality and integrity protection for messages thatendpoints exchange.¶

Internally, TLS is a layered protocol, with the structure shown inFigure 1.¶

          +-------------+------------+--------------+---------+Content   |             |            |  Application |         |Layer     |  Handshake  |   Alerts   |     Data     |   ...   |          |             |            |              |         |          +-------------+------------+--------------+---------+Record    |                                                   |Layer     |                      Records                      |          |                                                   |          +---------------------------------------------------+

Each content-layer message (e.g., handshake, alerts, and application data) iscarried as a series of typed TLS records by the record layer. Records areindividually cryptographically protected and then transmitted over a reliabletransport (typically TCP), which provides sequencing and guaranteed delivery.¶

The TLS authenticated key exchange occurs between two endpoints: client andserver. The client initiates the exchange and the server responds. If the keyexchange completes successfully, both client and server will agree on a secret.TLS supports both pre-shared key (PSK) and Diffie-Hellman over either finitefields or elliptic curves ((EC)DHE) key exchanges. PSK is the basis for EarlyData (0-RTT); the latter provides forward secrecy (FS) when the (EC)DHEkeys are destroyed. The two modes can also be combined to provide forwardsecrecy while using the PSK for authentication.¶

After completing the TLS handshake, the client will have learned andauthenticated an identity for the server, and the server is optionally able tolearn and authenticate an identity for the client. TLS supports X.509[RFC5280] certificate-based authentication for both server and client.When PSK key exchange is used (as in resumption), knowledge of the PSKserves to authenticate the peer.¶

The TLS key exchange is resistant to tampering by attackers, and it producesshared secrets that cannot be controlled by either participating peer.¶

TLS provides two basic handshake modes of interest to QUIC:¶

• A full 1-RTT handshake, in which the client is able to send application dataafter one round trip and the server immediately responds after receiving thefirst handshake message from the client.¶
• A 0-RTT handshake, in which the client uses information it has previouslylearned about the server to send application data immediately. Thisapplication data can be replayed by an attacker, so 0-RTT is not suitable forcarrying instructions that might initiate any action that could causeunwanted effects if replayed.¶

A simplified TLS handshake with 0-RTT application data is shown inFigure 2.¶

    Client                                             Server    ClientHello   (0-RTT Application Data)  -------->                                                  ServerHello                                         {EncryptedExtensions}                                                    {Finished}                             <--------      [Application Data]   {Finished}                -------->   [Application Data]        <------->      [Application Data]    () Indicates messages protected by Early Data (0-RTT) Keys    {} Indicates messages protected using Handshake Keys    [] Indicates messages protected using Application Data       (1-RTT) Keys

Figure 2 omits the EndOfEarlyData message, which is not used in QUIC; seeSection 8.3. Likewise, neither ChangeCipherSpec nor KeyUpdate messages areused by QUIC. ChangeCipherSpec is redundant in TLS 1.3; seeSection 8.4.QUIC has its own key update mechanism; seeSection 6.¶

Data is protected using a number of encryption levels:¶

• Initial keys¶
• Early data (0-RTT) keys¶
• Handshake keys¶
• Application data (1-RTT) keys¶

Application data can only appear in the early data and application datalevels. Handshake and alert messages may appear in any level.¶

The 0-RTT handshake can be used if the client and server have previouslycommunicated. In the 1-RTT handshake, the client is unable to send protectedapplication data until it has received all of the handshake messages sent by theserver.¶

4.1.Interface to TLS

As shown inFigure 4, the interface from QUIC to TLS consists of fourprimary functions:¶

• Sending and receiving handshake messages¶
• Processing stored transport and application state from a resumed sessionand determining if it is valid to generate or accept 0-RTT data¶
• Rekeying (both transmit and receive)¶
• Updating handshake state¶

Additional functions might be needed to configure TLS. In particular, QUIC andTLS need to agree on which is responsible for validation of peer credentials,such as certificate validation[RFC5280].¶

Client                                                    Server======                                                    ======Get Handshake                     Initial ------------->Install tx 0-RTT keys                     0-RTT - - - - - - - ->                                              Handshake Received                                                   Get Handshake                     <------------- Initial                                           Install rx 0-RTT keys                                          Install Handshake keys                                                   Get Handshake                     <----------- Handshake                                           Install tx 1-RTT keys                     <- - - - - - - - 1-RTTHandshake Received (Initial)Install Handshake keysHandshake Received (Handshake)Get Handshake                     Handshake ----------->Handshake CompleteInstall 1-RTT keys                     1-RTT - - - - - - - ->                                              Handshake Received                                              Handshake Complete                                             Handshake Confirmed                                           Install rx 1-RTT keys                     <--------------- 1-RTT                           (HANDSHAKE_DONE)Handshake Confirmed

4.2.TLS Version

This document describes how TLS 1.3[TLS13] is used with QUIC.¶

In practice, the TLS handshake will negotiate a version of TLS to use. Thiscould result in a version of TLS newer than 1.3 being negotiated if bothendpoints support that version. This is acceptable provided that the featuresof TLS 1.3 that are used by QUIC are supported by the newer version.¶

ClientsMUST NOT offer TLS versions older than 1.3. A badly configured TLSimplementation could negotiate TLS 1.2 or another older version of TLS. AnendpointMUST terminate the connection if a version of TLS older than 1.3 isnegotiated.¶

4.3.ClientHello Size

The first Initial packet from a client contains the start or all of its firstcryptographic handshake message, which for TLS is the ClientHello. Serversmight need to parse the entire ClientHello (e.g., to access extensions such asServer Name Identification (SNI) or Application-Layer Protocol Negotiation(ALPN)) in order to decide whether to accept the new incoming QUIC connection.If the ClientHello spans multiple Initial packets, such servers would need tobuffer the first received fragments, which could consume excessive resources ifthe client's address has not yet been validated. To avoid this, serversMAYuse the Retry feature (seeSection 8.1 of [QUIC-TRANSPORT]) to only bufferpartial ClientHello messages from clients with a validated address.¶

QUIC packet and framing add at least 36 bytes of overhead to the ClientHellomessage. That overhead increases if the client chooses a Source Connection IDfield longer than zero bytes. Overheads also do not include the token or aDestination Connection ID longer than 8 bytes, both of which might be requiredif a server sends a Retry packet.¶

A typical TLS ClientHello can easily fit into a 1200-byte packet. However, inaddition to the overheads added by QUIC, there are several variables that couldcause this limit to be exceeded. Large session tickets, multiple or large keyshares, and long lists of supported ciphers, signature algorithms, versions,QUIC transport parameters, and other negotiable parameters and extensions couldcause this message to grow.¶

For servers, in addition to connection IDs and tokens, the size of TLS sessiontickets can have an effect on a client's ability to connect efficiently.Minimizing the size of these values increases the probability that clients canuse them and still fit their entire ClientHello message in their first Initialpacket.¶

The TLS implementation does not need to ensure that the ClientHello is largeenough to meet QUIC's requirements for datagrams that carry Initial packets; seeSection 14.1 of [QUIC-TRANSPORT]. QUIC implementations use PADDING frames orpacket coalescing to ensure that datagrams are large enough.¶

4.4.Peer Authentication

The requirements for authentication depend on the application protocol that isin use. TLS provides server authentication and permits the server to requestclient authentication.¶

A clientMUST authenticate the identity of the server. This typically involvesverification that the identity of the server is included in a certificate andthat the certificate is issued by a trusted entity (see for example[RFC2818]).¶

Note: Where servers provide certificates for authentication, the size of the certificate chain can consume a large number of bytes. Controlling the size of certificate chains is critical to performance in QUIC as servers are limited to sending 3 bytes for every byte received prior to validating the client address; seeSection 8.1 of [QUIC-TRANSPORT]. The size of a certificate chain can be managed by limiting the number of names or extensions; using keys with small public key representations, like ECDSA; or by using certificate compression[COMPRESS].¶

A serverMAY request that the client authenticate during the handshake. A serverMAY refuse a connection if the client is unable to authenticate when requested.The requirements for client authentication vary based on application protocoland deployment.¶

A serverMUST NOT use post-handshake client authentication (as defined inSection 4.6.2 of [TLS13]) because the multiplexing offered by QUIC preventsclients from correlating the certificate request with the application-levelevent that triggered it (see[HTTP2-TLS13]). More specifically,serversMUST NOT send post-handshake TLS CertificateRequest messages, andclientsMUST treat receipt of such messages as a connection error of typePROTOCOL_VIOLATION.¶

4.5.Session Resumption

QUIC can use the session resumption feature of TLS 1.3. It does this bycarrying NewSessionTicket messages in CRYPTO frames after the handshake iscomplete. Session resumption can be used to provide 0-RTT and can also beused when 0-RTT is disabled.¶

Endpoints that use session resumption might need to remember some informationabout the current connection when creating a resumed connection. TLS requiresthat some information be retained; seeSection 4.6.1 of [TLS13]. QUIC itselfdoes not depend on any state being retained when resuming a connection unless0-RTT is also used; seeSection 7.4.1 of [QUIC-TRANSPORT] andSection 4.6.1. Application protocols could depend on state that is retainedbetween resumed connections.¶

Clients can store any state required for resumption along with the sessionticket. Servers can use the session ticket to help carry state.¶

Session resumption allows servers to link activity on the original connectionwith the resumed connection, which might be a privacy issue for clients.Clients can choose not to enable resumption to avoid creating this correlation.ClientsSHOULD NOT reuse tickets as that allows entities other than the serverto correlate connections; seeAppendix C.4 of [TLS13].¶

4.6.0-RTT

The 0-RTT feature in QUIC allows a client to send application data before thehandshake is complete. This is made possible by reusing negotiated parametersfrom a previous connection. To enable this, 0-RTT depends on the clientremembering critical parameters and providing the server with a TLS sessionticket that allows the server to recover the same information.¶

This information includes parameters that determine TLS state, as governed by[TLS13], QUIC transport parameters, the chosen application protocol, and anyinformation the application protocol might need; seeSection 4.6.3. Thisinformation determines how 0-RTT packets and their contents are formed.¶

To ensure that the same information is available to both endpoints, allinformation used to establish 0-RTT comes from the same connection. Endpointscannot selectively disregard information that might alter the sending orprocessing of 0-RTT.¶

[TLS13] sets a limit of seven days on the time between the originalconnection and any attempt to use 0-RTT. There are other constraints on 0-RTTusage, notably those caused by the potential exposure to replay attack; seeSection 9.2.¶

4.7.HelloRetryRequest

The HelloRetryRequest message (seeSection 4.1.4 of [TLS13]) can be used torequest that a client provide new information, such as a key share, or tovalidate some characteristic of the client. From the perspective of QUIC,HelloRetryRequest is not differentiated from other cryptographic handshakemessages that are carried in Initial packets. Although it is in principlepossible to use this feature for address verification, QUIC implementationsSHOULD instead use the Retry feature; seeSection 8.1 of [QUIC-TRANSPORT].¶

4.8.TLS Errors

If TLS experiences an error, it generates an appropriate alert as defined inSection 6 of [TLS13].¶

A TLS alert is converted into a QUIC connection error. The AlertDescriptionvalue isadded to 0x0100 to produce a QUIC error code from the range reserved forCRYPTO_ERROR; seeSection 20.1 of [QUIC-TRANSPORT]. The resulting value issent in a QUIC CONNECTION_CLOSE frame of type 0x1c.¶

QUIC is only able to convey an alert level of "fatal". In TLS 1.3, the onlyexisting uses for the "warning" level are to signal connection close; seeSection 6.1 of [TLS13]. As QUIC provides alternative mechanisms forconnection termination and the TLS connection is only closed if an error isencountered, a QUIC endpointMUST treat any alert from TLS as if it were at the"fatal" level.¶

QUIC permits the use of a generic code in place of a specific error code; seeSection 11 of [QUIC-TRANSPORT]. For TLS alerts, this includes replacing anyalert with a generic alert, such as handshake_failure (0x0128 in QUIC).EndpointsMAY use a generic error code to avoid possibly exposing confidentialinformation.¶

4.9.Discarding Unused Keys

After QUIC has completed a move to a new encryption level, packet protectionkeys for previous encryption levels can be discarded. This occurs several timesduring the handshake, as well as when keys are updated; seeSection 6.¶

Packet protection keys are not discarded immediately when new keys areavailable. If packets from a lower encryption level contain CRYPTO frames,frames that retransmit that dataMUST be sent at the same encryption level.Similarly, an endpoint generates acknowledgments for packets at the sameencryption level as the packet being acknowledged. Thus, it is possible thatkeys for a lower encryption level are needed for a short time after keys for anewer encryption level are available.¶

An endpoint cannot discard keys for a given encryption level unless it hasreceived all the cryptographic handshake messages from its peer at thatencryption level and its peer has done the same. Different methods fordetermining this are provided for Initial keys (Section 4.9.1) andHandshake keys (Section 4.9.2). These methods do not prevent packetsfrom being received or sent at that encryption level because a peer might nothave received all the acknowledgments necessary.¶

Though an endpoint might retain older keys, new dataMUST be sent at the highestcurrently available encryption level. Only ACK frames and retransmissions ofdata in CRYPTO frames are sent at a previous encryption level. These packetsMAY also include PADDING frames.¶

5.1.Packet Protection Keys

QUIC derives packet protection keys in the same way that TLS derives recordprotection keys.¶

Each encryption level has separate secret values for protection of packets sentin each direction. These traffic secrets are derived by TLS (seeSection 7.1 of [TLS13]) and are used by QUIC for all encryption levels except the Initialencryption level. The secrets for the Initial encryption level are computedbased on the client's initial Destination Connection ID, as described inSection 5.2.¶

The keys used for packet protection are computed from the TLS secrets using theKDF provided by TLS. In TLS 1.3, the HKDF-Expand-Label function described inSection 7.1 of [TLS13] is used with the hash function from the negotiatedcipher suite. All uses of HKDF-Expand-Label in QUIC use a zero-length Context.¶

Note that labels, which are described using strings, are encodedas bytes using ASCII[ASCII] without quotes or any trailing NULbyte.¶

Other versions of TLSMUST provide a similar function in order to beused with QUIC.¶

The current encryption level secret and the label "quic key" are input to theKDF to produce the AEAD key; the label "quic iv" is used to derive theInitialization Vector (IV); seeSection 5.3. The header protection key uses the"quic hp" label; seeSection 5.4. Using these labels provides keyseparation between QUIC and TLS; seeSection 9.6.¶

Both "quic key" and "quic hp" are used to produce keys, so the Length providedto HKDF-Expand-Label along with these labels is determined by the size of keysin the AEAD or header protection algorithm. The Length provided with "quic iv"is the minimum length of the AEAD nonce or 8 bytes if that is larger; see[AEAD].¶

The KDF used for initial secrets is always the HKDF-Expand-Label function fromTLS 1.3; seeSection 5.2.¶

5.2.Initial Secrets

Initial packets apply the packet protection process, but use a secret derivedfrom the Destination Connection ID field from the client's first Initialpacket.¶

This secret is determined by using HKDF-Extract (seeSection 2.2 of [HKDF])with a salt of 0x38762cf7f55934b34d179ae6a4c80cadccbb7f0a and the input keyingmaterial (IKM) of the Destination Connection ID field. This produces anintermediate pseudorandom key (PRK) that is used to derive two separate secretsfor sending and receiving.¶

The secret used by clients to construct Initial packets uses the PRK and thelabel "client in" as input to the HKDF-Expand-Label function from TLS[TLS13] to produce a 32-byte secret. Packets constructed by the server usethe same process with the label "server in". The hash function for HKDF whenderiving initial secrets and keys is SHA-256[SHA].¶

This process in pseudocode is:¶

initial_salt = 0x38762cf7f55934b34d179ae6a4c80cadccbb7f0ainitial_secret = HKDF-Extract(initial_salt,                              client_dst_connection_id)client_initial_secret = HKDF-Expand-Label(initial_secret,                                          "client in", "",                                          Hash.length)server_initial_secret = HKDF-Expand-Label(initial_secret,                                          "server in", "",                                          Hash.length)

The connection ID used with HKDF-Expand-Label is the Destination Connection IDin the Initial packet sent by the client. This will be a randomly selectedvalue unless the client creates the Initial packet after receiving a Retrypacket, where the Destination Connection ID is selected by the server.¶

Future versions of QUICSHOULD generate a new salt value, thus ensuring thatthe keys are different for each version of QUIC. This prevents a middlebox thatrecognizes only one version of QUIC from seeing or modifying the contents ofpackets from future versions.¶

The HKDF-Expand-Label function defined in TLS 1.3MUST be used for Initialpackets even where the TLS versions offered do not include TLS 1.3.¶

The secrets used for constructing subsequent Initial packets change when aserver sends a Retry packet to use the connection ID value selected by theserver. The secrets do not change when a client changes the DestinationConnection ID it uses in response to an Initial packet from the server.¶

Note: The Destination Connection ID field could be any length up to 20 bytes, including zero length if the server sends a Retry packet with a zero-length Source Connection ID field. After a Retry, the Initial keys provide the client no assurance that the server received its packet, so the client has to rely on the exchange that included the Retry packet to validate the server address; seeSection 8.1 of [QUIC-TRANSPORT].¶

Appendix A contains sample Initial packets.¶

5.3.AEAD Usage

The Authenticated Encryption with Associated Data (AEAD) function (see[AEAD]) used for QUIC packet protection is the AEAD that is negotiated foruse with the TLS connection. For example, if TLS is using theTLS_AES_128_GCM_SHA256 cipher suite, the AEAD_AES_128_GCM function is used.¶

QUIC can use any of the cipher suites defined in[TLS13] with the exceptionof TLS_AES_128_CCM_8_SHA256. A cipher suiteMUST NOT be negotiated unless aheader protection scheme is defined for the cipher suite. This document definesa header protection scheme for all cipher suites defined in[TLS13] asidefrom TLS_AES_128_CCM_8_SHA256. These cipher suites have a 16-byteauthentication tag and produce an output 16 bytes larger than their input.¶

An endpointMUST NOT reject a ClientHello that offers a cipher suite that itdoes not support, or it would be impossible to deploy a new cipher suite. Thisalso applies to TLS_AES_128_CCM_8_SHA256.¶

When constructing packets, the AEAD function is applied prior to applyingheader protection; seeSection 5.4. The unprotected packet header is partof the associated data (A). When processing packets, an endpoint firstremoves the header protection.¶

The key and IV for the packet are computed as described inSection 5.1.The nonce, N, is formed by combining the packet protection IV with the packetnumber. The 62 bits of the reconstructed QUIC packet number in network byteorder are left-padded with zeros to the size of the IV. The exclusive OR of thepadded packet number and the IV forms the AEAD nonce.¶

The associated data, A, for the AEAD is the contents of the QUIC header,starting from the first byte of either the short or long header, up to andincluding the unprotected packet number.¶

The input plaintext, P, for the AEAD is the payload of the QUIC packet, asdescribed in[QUIC-TRANSPORT].¶

The output ciphertext, C, of the AEAD is transmitted in place of P.¶

Some AEAD functions have limits for how many packets can be encrypted under thesame key and IV; seeSection 6.6. This might be lower than the packetnumber limit. An endpointMUST initiate a key update (Section 6) prior toexceeding any limit set for the AEAD that is in use.¶

5.4.Header Protection

Parts of QUIC packet headers, in particular the Packet Number field, areprotected using a key that is derived separately from the packet protection keyand IV. The key derived using the "quic hp" label is used to provideconfidentiality protection for those fields that are not exposed to on-pathelements.¶

This protection applies to the least significant bits of the first byte, plusthe Packet Number field. The four least significant bits of the first byte areprotected for packets with long headers; the five least significant bits of thefirst byte are protected for packets with short headers. For both header forms,this covers the reserved bits and the Packet Number Length field; the Key Phasebit is also protected for packets with a short header.¶

The same header protection key is used for the duration of the connection, withthe value not changing after a key update (seeSection 6). This allowsheader protection to be used to protect the key phase.¶

This process does not apply to Retry or Version Negotiation packets, which donot contain a protected payload or any of the fields that are protected by thisprocess.¶

mask = header_protection(hp_key, sample)pn_length = (packet[0] & 0x03) + 1if (packet[0] & 0x80) == 0x80:   # Long header: 4 bits masked   packet[0] ^= mask[0] & 0x0felse:   # Short header: 5 bits masked   packet[0] ^= mask[0] & 0x1f# pn_offset is the start of the Packet Number field.packet[pn_offset:pn_offset+pn_length] ^= mask[1:1+pn_length]

Initial Packet {  Header Form (1) = 1,  Fixed Bit (1) = 1,  Long Packet Type (2) = 0,  Reserved Bits (2),         # Protected  Packet Number Length (2),  # Protected  Version (32),  DCID Len (8),  Destination Connection ID (0..160),  SCID Len (8),  Source Connection ID (0..160),  Token Length (i),  Token (..),  Length (i),  Packet Number (8..32),     # Protected  Protected Payload (0..24), # Skipped Part  Protected Payload (128),   # Sampled Part  Protected Payload (..)     # Remainder}1-RTT Packet {  Header Form (1) = 0,  Fixed Bit (1) = 1,  Spin Bit (1),  Reserved Bits (2),         # Protected  Key Phase (1),             # Protected  Packet Number Length (2),  # Protected  Destination Connection ID (0..160),  Packet Number (8..32),     # Protected  Protected Payload (0..24), # Skipped Part  Protected Payload (128),   # Sampled Part  Protected Payload (..),    # Remainder}

# pn_offset is the start of the Packet Number field.sample_offset = pn_offset + 4sample = packet[sample_offset..sample_offset+sample_length]

pn_offset = 1 + len(connection_id)

pn_offset = 7 + len(destination_connection_id) +                len(source_connection_id) +                len(payload_length)if packet_type == Initial:    pn_offset += len(token_length) +                 len(token)

header_protection(hp_key, sample):  mask = AES-ECB(hp_key, sample)

header_protection(hp_key, sample):  counter = sample[0..3]  nonce = sample[4..15]  mask = ChaCha20(hp_key, counter, nonce, {0,0,0,0,0})

5.5.Receiving Protected Packets

Once an endpoint successfully receives a packet with a given packet number, itMUST discard all packets in the same packet number space with higher packetnumbers if they cannot be successfully unprotected with either the same key, or-- if there is a key update -- a subsequent packet protection key; seeSection 6. Similarly, a packet that appears to trigger a key update butcannot be unprotected successfullyMUST be discarded.¶

Failure to unprotect a packet does not necessarily indicate the existence of aprotocol error in a peer or an attack. The truncated packet number encodingused in QUIC can cause packet numbers to be decoded incorrectly if they aredelayed significantly.¶

5.6.Use of 0-RTT Keys

If 0-RTT keys are available (seeSection 4.6.1), the lack of replay protectionmeans that restrictions on their use are necessary to avoid replay attacks onthe protocol.¶

Of the frames defined in[QUIC-TRANSPORT], the STREAM, RESET_STREAM,STOP_SENDING, and CONNECTION_CLOSE frames are potentially unsafe for use with0-RTT as they carry application data. Application data that is received in0-RTT could cause an application at the server to process the data multipletimes rather than just once. Additional actions taken by a server as a resultof processing replayed application data could have unwanted consequences. Aclient thereforeMUST NOT use 0-RTT for application data unless specificallyrequested by the application that is in use.¶

An application protocol that uses QUICMUST include a profile that definesacceptable use of 0-RTT; otherwise, 0-RTT can only be used to carry QUIC framesthat do not carry application data. For example, a profile for HTTP isdescribed in[HTTP-REPLAY] and used for HTTP/3; seeSection 10.9 of [QUIC-HTTP].¶

Though replaying packets might result in additional connection attempts, theeffect of processing replayed frames that do not carry application data islimited to changing the state of the affected connection. A TLS handshakecannot be successfully completed using replayed packets.¶

A clientMAY wish to apply additional restrictions on what data it sends priorto the completion of the TLS handshake.¶

A client otherwise treats 0-RTT keys as equivalent to 1-RTT keys, except thatit cannot send certain frames with 0-RTT keys; seeSection 12.5 of [QUIC-TRANSPORT].¶

A client that receives an indication that its 0-RTT data has been accepted by aserver can send 0-RTT data until it receives all of the server's handshakemessages. A clientSHOULD stop sending 0-RTT data if it receives an indicationthat 0-RTT data has been rejected.¶

A serverMUST NOT use 0-RTT keys to protect packets; it uses 1-RTT keys toprotect acknowledgments of 0-RTT packets. A clientMUST NOT attempt todecrypt 0-RTT packets it receives and insteadMUST discard them.¶

Once a client has installed 1-RTT keys, itMUST NOT send any more 0-RTTpackets.¶

Note: 0-RTT data can be acknowledged by the server as it receives it, but any packets containing acknowledgments of 0-RTT data cannot have packet protection removed by the client until the TLS handshake is complete. The 1-RTT keys necessary to remove packet protection cannot be derived until the client receives all server handshake messages.¶

5.7.Receiving Out-of-Order Protected Packets

Due to reordering and loss, protected packets might be received by an endpointbefore the final TLS handshake messages are received. A client will be unableto decrypt 1-RTT packets from the server, whereas a server will be able todecrypt 1-RTT packets from the client. Endpoints in either roleMUST NOTdecrypt 1-RTT packets from their peer prior to completing the handshake.¶

Even though 1-RTT keys are available to a server after receiving the firsthandshake messages from a client, it is missing assurances on the client state:¶

• The client is not authenticated, unless the server has chosen to use apre-shared key and validated the client's pre-shared key binder; seeSection 4.2.11 of [TLS13].¶
• The client has not demonstrated liveness, unless the server has validated theclient's address with a Retry packet or other means; seeSection 8.1 of [QUIC-TRANSPORT].¶
• Any received 0-RTT data that the server responds to might be due to a replayattack.¶

Therefore, the server's use of 1-RTT keys before the handshake is complete islimited to sending data. A serverMUST NOT process incoming 1-RTT protectedpackets before the TLS handshake is complete. Because sending acknowledgmentsindicates that all frames in a packet have been processed, a server cannot sendacknowledgments for 1-RTT packets until the TLS handshake is complete. Receivedpackets protected with 1-RTT keysMAY be stored and later decrypted and usedonce the handshake is complete.¶

Note: TLS implementations might provide all 1-RTT secrets prior to handshake completion. Even where QUIC implementations have 1-RTT read keys, those keys are not to be used prior to completing the handshake.¶

The requirement for the server to wait for the client Finished message createsa dependency on that message being delivered. A client can avoid thepotential for head-of-line blocking that this implies by sending its 1-RTTpackets coalesced with a Handshake packet containing a copy of the CRYPTO framethat carries the Finished message, until one of the Handshake packets isacknowledged. This enables immediate server processing for those packets.¶

A server could receive packets protected with 0-RTT keys prior to receiving aTLS ClientHello. The serverMAY retain these packets for later decryption inanticipation of receiving a ClientHello.¶

A client generally receives 1-RTT keys at the same time as the handshakecompletes. Even if it has 1-RTT secrets, a clientMUST NOT processincoming 1-RTT protected packets before the TLS handshake is complete.¶

5.8.Retry Packet Integrity

Retry packets (seeSection 17.2.5 of [QUIC-TRANSPORT]) carry a Retry IntegrityTag that provides two properties: it allows the discarding of packets that haveaccidentally been corrupted by the network, and only an entity that observes anInitial packet can send a valid Retry packet.¶

The Retry Integrity Tag is a 128-bit field that is computed as the output ofAEAD_AES_128_GCM[AEAD] used with the following inputs:¶

• The secret key, K, is 128 bits equal to 0xbe0c690b9f66575a1d766b54e368c84e.¶
• The nonce, N, is 96 bits equal to 0x461599d35d632bf2239825bb.¶
• The plaintext, P, is empty.¶
• The associated data, A, is the contents of the Retry Pseudo-Packet, asillustrated inFigure 8:¶

The secret key and the nonce are values derived by calling HKDF-Expand-Labelusing 0xd9c9943e6101fd200021506bcc02814c73030f25c79d71ce876eca876e6fca8e as thesecret, with labels being "quic key" and "quic iv" (Section 5.1).¶

Retry Pseudo-Packet {  ODCID Length (8),  Original Destination Connection ID (0..160),  Header Form (1) = 1,  Fixed Bit (1) = 1,  Long Packet Type (2) = 3,  Unused (4),  Version (32),  DCID Len (8),  Destination Connection ID (0..160),  SCID Len (8),  Source Connection ID (0..160),  Retry Token (..),}

The Retry Pseudo-Packet is not sent over the wire. It is computed by takingthe transmitted Retry packet, removing the Retry Integrity Tag, and prependingthe two following fields:¶

ODCID Length:

The ODCID Length field contains the length in bytes of the OriginalDestination Connection ID field that follows it, encoded as an 8-bit unsignedinteger.¶

Original Destination Connection ID:

The Original Destination Connection ID contains the value of the DestinationConnection ID from the Initial packet that this Retry is in response to. Thelength of this field is given in ODCID Length. The presence of this fieldensures that a valid Retry packet can only be sent by an entity thatobserves the Initial packet.¶

6.1.Initiating a Key Update

Endpoints maintain separate read and write secrets for packet protection. Anendpoint initiates a key update by updating its packet protection write secretand using that to protect new packets. The endpoint creates a new write secretfrom the existing write secret as performed inSection 7.2 of [TLS13]. Thisuses the KDF function provided by TLS with a label of "quic ku". Thecorresponding key and IV are created from that secret as defined inSection 5.1. The header protection key is not updated.¶

For example, to update write keys with TLS 1.3, HKDF-Expand-Label is used as:¶

secret_<n+1> = HKDF-Expand-Label(secret_<n>, "quic ku",                                 "", Hash.length)

The endpoint toggles the value of the Key Phase bit and uses the updated key andIV to protect all subsequent packets.¶

An endpointMUST NOT initiate a key update prior to having confirmed thehandshake (Section 4.1.2). An endpointMUST NOT initiate a subsequentkey update unless it has received an acknowledgment for a packet that was sentprotected with keys from the current key phase. This ensures that keys areavailable to both peers before another key update can be initiated. This can beimplemented by tracking the lowest packet number sent with each key phase andthe highest acknowledged packet number in the 1-RTT space: once the latter ishigher than or equal to the former, another key update can be initiated.¶

Note: Keys of packets other than the 1-RTT packets are never updated; their keys are derived solely from the TLS handshake state.¶

The endpoint that initiates a key update also updates the keys that it uses forreceiving packets. These keys will be needed to process packets the peer sendsafter updating.¶

An endpointMUST retain old keys until it has successfully unprotected a packetsent using the new keys. An endpointSHOULD retain old keys for some timeafter unprotecting a packet sent using the new keys. Discarding old keys tooearly can cause delayed packets to be discarded. Discarding packets will beinterpreted as packet loss by the peer and could adversely affect performance.¶

6.2.Responding to a Key Update

A peer is permitted to initiate a key update after receiving an acknowledgmentof a packet in the current key phase. An endpoint detects a key update whenprocessing a packet with a key phase that differs from the value used to protectthe last packet it sent. To process this packet, the endpoint uses the nextpacket protection key and IV. SeeSection 6.3 for considerationsabout generating these keys.¶

If a packet is successfully processed using the next key and IV, then the peerhas initiated a key update. The endpointMUST update its send keys to thecorresponding key phase in response, as described inSection 6.1.Sending keysMUST be updated before sending an acknowledgment for the packetthat was received with updated keys. By acknowledging the packet that triggeredthe key update in a packet protected with the updated keys, the endpoint signalsthat the key update is complete.¶

An endpoint can defer sending the packet or acknowledgment according to itsnormal packet sending behavior; it is not necessary to immediately generate apacket in response to a key update. The next packet sent by the endpoint willuse the updated keys. The next packet that contains an acknowledgment willcause the key update to be completed. If an endpoint detects a second updatebefore it has sent any packets with updated keys containing anacknowledgment for the packet that initiated the key update, it indicates thatits peer has updated keys twice without awaiting confirmation. An endpointMAYtreat such consecutive key updates as a connection error of typeKEY_UPDATE_ERROR.¶

An endpoint that receives an acknowledgment that is carried in a packetprotected with old keys where any acknowledged packet was protected with newerkeysMAY treat that as a connection error of type KEY_UPDATE_ERROR. Thisindicates that a peer has received and acknowledged a packet that initiates akey update, but has not updated keys in response.¶

6.3.Timing of Receive Key Generation

Endpoints responding to an apparent key updateMUST NOT generate a timingside-channel signal that might indicate that the Key Phase bit was invalid (seeSection 9.5). Endpoints can use randomized packet protection keys inplace of discarded keys when key updates are not yet permitted. Usingrandomized keys ensures that attempting to remove packet protection does notresult in timing variations, and results in packets with an invalid Key Phasebit being rejected.¶

The process of creating new packet protection keys for receiving packets couldreveal that a key update has occurred. An endpointMAY generate new keys aspart of packet processing, but this creates a timing signal that could be usedby an attacker to learn when key updates happen and thus leak the value of theKey Phase bit.¶

Endpoints are generally expected to have current and next receive packetprotection keys available. For a short period after a key update completes, upto the PTO, endpointsMAY defer generation of the next set ofreceive packet protection keys. This allows endpointsto retain only two sets of receive keys; seeSection 6.5.¶

Once generated, the next set of packet protection keysSHOULD be retained, evenif the packet that was received was subsequently discarded. Packets containingapparent key updates are easy to forge, and while the process of key update doesnot require significant effort, triggering this process could be used by anattacker for DoS.¶

For this reason, endpointsMUST be able to retain two sets of packet protectionkeys for receiving packets: the current and the next. Retaining the previouskeys in addition to these might improve performance, but this is not essential.¶

6.4.Sending with Updated Keys

An endpoint never sends packets that are protected with old keys. Only thecurrent keys are used. Keys used for protecting packets can be discardedimmediately after switching to newer keys.¶

Packets with higher packet numbersMUST be protected with either the same ornewer packet protection keys than packets with lower packet numbers. Anendpoint that successfully removes protection with old keys when newer keys wereused for packets with lower packet numbersMUST treat this as a connection errorof type KEY_UPDATE_ERROR.¶

6.5.Receiving with Different Keys

For receiving packets during a key update, packets protected with older keysmight arrive if they were delayed by the network. Retaining old packetprotection keys allows these packets to be successfully processed.¶

As packets protected with keys from the next key phase use the same Key Phasevalue as those protected with keys from the previous key phase, it is necessaryto distinguish between the two if packets protected with old keys are to beprocessed. This can be done using packet numbers. A recovered packet numberthat is lower than any packet number from the current key phase uses theprevious packet protection keys; a recovered packet number that is higher thanany packet number from the current key phase requires the use of the next packetprotection keys.¶

Some care is necessary to ensure that any process for selecting betweenprevious, current, and next packet protection keys does not expose a timing sidechannel that might reveal which keys were used to remove packet protection. SeeSection 9.5 for more information.¶

Alternatively, endpoints can retain only two sets of packet protection keys,swapping previous for next after enough time has passed to allow for reorderingin the network. In this case, the Key Phase bit alone can be used to selectkeys.¶

An endpointMAY allow a period of approximately the Probe Timeout (PTO; see[QUIC-RECOVERY]) after promoting the next set of receive keys to be currentbefore it creates the subsequent set of packet protection keys. These updatedkeysMAY replace the previous keys at that time. With the caveat that PTO is asubjective measure -- that is, a peer could have a different view of the RTT --this time is expected to be long enough that any reordered packets would bedeclared lost by a peer even if they were acknowledged and short enough to allowa peer to initiate further key updates.¶

Endpoints need to allow for the possibility that a peer might not be able todecrypt packets that initiate a key update during the period when the peerretains old keys. EndpointsSHOULD wait three times the PTO before initiating akey update after receiving an acknowledgment that confirms that the previous keyupdate was received. Failing to allow sufficient time could lead to packetsbeing discarded.¶

An endpointSHOULD retain old read keys for no more than three times the PTOafter having received a packet protected using the new keys. After this period,old read keys and their corresponding secretsSHOULD be discarded.¶

6.6.Limits on AEAD Usage

This document sets usage limits for AEAD algorithms to ensure that overuse doesnot give an adversary a disproportionate advantage in attacking theconfidentiality and integrity of communications when using QUIC.¶

The usage limits defined in TLS 1.3 exist for protection against attackson confidentiality and apply to successful applications of AEAD protection. Theintegrity protections in authenticated encryption also depend on limiting thenumber of attempts to forge packets. TLS achieves this by closing connectionsafter any record fails an authentication check. In comparison, QUIC ignores anypacket that cannot be authenticated, allowing multiple forgery attempts.¶

QUIC accounts for AEAD confidentiality and integrity limits separately. Theconfidentiality limit applies to the number of packets encrypted with a givenkey. The integrity limit applies to the number of packets decrypted within agiven connection. Details on enforcing these limits for each AEAD algorithmfollow below.¶

EndpointsMUST count the number of encrypted packets for each set of keys. Ifthe total number of encrypted packets with the same key exceeds theconfidentiality limit for the selected AEAD, the endpointMUST stop using thosekeys. EndpointsMUST initiate a key update before sending more protected packetsthan the confidentiality limit for the selected AEAD permits. If a key updateis not possible or integrity limits are reached, the endpointMUST stop usingthe connection and only send stateless resets in response to receiving packets.It isRECOMMENDED that endpoints immediately close the connection with aconnection error of type AEAD_LIMIT_REACHED before reaching a state where keyupdates are not possible.¶

For AEAD_AES_128_GCM and AEAD_AES_256_GCM, the confidentiality limit is2²³ encrypted packets; seeAppendix B.1. ForAEAD_CHACHA20_POLY1305, the confidentiality limit is greater than the number ofpossible packets (2⁶²) and so can be disregarded. ForAEAD_AES_128_CCM, the confidentiality limit is 2^21.5 encryptedpackets; seeAppendix B.2. Applying a limit reduces the probability that anattacker can distinguish the AEAD in use from a random permutation; see[AEBounds],[ROBUST], and[GCM-MU].¶

In addition to counting packets sent, endpointsMUST count the number ofreceived packets that fail authentication during the lifetime of a connection.If the total number of received packets that fail authentication within theconnection, across all keys, exceeds the integrity limit for the selected AEAD,the endpointMUST immediately close the connection with a connection error oftype AEAD_LIMIT_REACHED and not process any more packets.¶

For AEAD_AES_128_GCM and AEAD_AES_256_GCM, the integrity limit is 2⁵²invalid packets; seeAppendix B.1. For AEAD_CHACHA20_POLY1305, the integritylimit is 2³⁶ invalid packets; see[AEBounds]. For AEAD_AES_128_CCM,the integrity limit is 2^21.5 invalid packets; seeAppendix B.2. Applying this limit reduces the probability that an attacker cansuccessfully forge a packet; see[AEBounds],[ROBUST], and[GCM-MU].¶

Endpoints that limit the size of packetsMAY use higher confidentiality andintegrity limits; seeAppendix B for details.¶

Future analyses and specificationsMAY relax confidentiality or integrity limitsfor an AEAD.¶

Any TLS cipher suite that is specified for use with QUICMUST define limits onthe use of the associated AEAD function that preserves margins forconfidentiality and integrity. That is, limitsMUST be specified for the numberof packets that can be authenticated and for the number of packets that can failauthentication. Providing a reference to any analysis upon which values arebased -- and any assumptions used in that analysis -- allows limits to beadapted to varying usage conditions.¶

6.7.Key Update Error Code

The KEY_UPDATE_ERROR error code (0x0e) is used to signal errors related to keyupdates.¶

8.1.Protocol Negotiation

QUIC requires that the cryptographic handshake provide authenticated protocolnegotiation. TLS uses Application-Layer Protocol Negotiation[ALPN] to select an application protocol. Unless another mechanismis used for agreeing on an application protocol, endpointsMUST use ALPN forthis purpose.¶

When using ALPN, endpointsMUST immediately close a connection (seeSection 10.2 of [QUIC-TRANSPORT]) with a no_application_protocol TLS alert (QUIC errorcode 0x0178; seeSection 4.8) if an application protocol is not negotiated.While[ALPN] only specifies that servers use this alert, QUIC clientsMUSTuse error 0x0178 to terminate a connection when ALPN negotiation fails.¶

An application protocolMAY restrict the QUIC versions that it can operate over.ServersMUST select an application protocol compatible with the QUIC versionthat the client has selected. The serverMUST treat the inability to select acompatible application protocol as a connection error of type 0x0178(no_application_protocol). Similarly, a clientMUST treat the selection of anincompatible application protocol by a server as a connection error of type0x0178.¶

8.2.QUIC Transport Parameters Extension

QUIC transport parameters are carried in a TLS extension. Different versions ofQUIC might define a different method for negotiating transport configuration.¶

Including transport parameters in the TLS handshake provides integrityprotection for these values.¶

   enum {      quic_transport_parameters(0x39), (65535)   } ExtensionType;

The extension_data field of the quic_transport_parameters extension contains avalue that is defined by the version of QUIC that is in use.¶

The quic_transport_parameters extension is carried in the ClientHello and theEncryptedExtensions messages during the handshake. EndpointsMUST send thequic_transport_parameters extension; endpoints that receive ClientHello orEncryptedExtensions messages without the quic_transport_parameters extensionMUST close the connection with an error of type 0x016d (equivalent to a fatalTLS missing_extension alert, seeSection 4.8).¶

Transport parameters become available prior to the completion of the handshake.A server might use these values earlier than handshake completion. However, thevalue of transport parameters is not authenticated until the handshakecompletes, so any use of these parameters cannot depend on their authenticity.Any tampering with transport parameters will cause the handshake to fail.¶

EndpointsMUST NOT send this extension in a TLS connection that does not useQUIC (such as the use of TLS with TCP defined in[TLS13]). A fatalunsupported_extension alertMUST be sent by an implementation that supports thisextension if the extension is received when the transport is not QUIC.¶

Negotiating the quic_transport_parameters extension causes the EndOfEarlyData tobe removed; seeSection 8.3.¶

8.3.Removing the EndOfEarlyData Message

The TLS EndOfEarlyData message is not used with QUIC. QUIC does not rely onthis message to mark the end of 0-RTT data or to signal the change to Handshakekeys.¶

ClientsMUST NOT send the EndOfEarlyData message. A serverMUST treat receiptof a CRYPTO frame in a 0-RTT packet as a connection error of typePROTOCOL_VIOLATION.¶

As a result, EndOfEarlyData does not appear in the TLS handshake transcript.¶

8.4.Prohibit TLS Middlebox Compatibility Mode

Appendix D.4 of[TLS13] describes an alteration to the TLS 1.3 handshake asa workaround for bugs in some middleboxes. The TLS 1.3 middlebox compatibilitymode involves setting the legacy_session_id field to a 32-byte value in theClientHello and ServerHello, then sending a change_cipher_spec record. Bothfield and record carry no semantic content and are ignored.¶

This mode has no use in QUIC as it only applies to middleboxes that interferewith TLS over TCP. QUIC also provides no means to carry a change_cipher_specrecord. A clientMUST NOT request the use of the TLS 1.3 compatibility mode. AserverSHOULD treat the receipt of a TLS ClientHello with a non-emptylegacy_session_id field as a connection error of type PROTOCOL_VIOLATION.¶

9.1.Session Linkability

Use of TLS session tickets allows servers and possibly other entities tocorrelate connections made by the same client; seeSection 4.5 for details.¶

9.2.Replay Attacks with 0-RTT

As described inSection 8 of [TLS13], use of TLS early data comes with anexposure to replay attack. The use of 0-RTT in QUIC is similarly vulnerable toreplay attack.¶

EndpointsMUST implement and use the replay protections described in[TLS13],however it is recognized that these protections are imperfect. Therefore,additional consideration of the risk of replay is needed.¶

QUIC is not vulnerable to replay attack, except via the application protocolinformation it might carry. The management of QUIC protocol state based on theframe types defined in[QUIC-TRANSPORT] is not vulnerable to replay.Processing of QUIC frames is idempotent and cannot result in invalid connectionstates if frames are replayed, reordered, or lost. QUIC connections do notproduce effects that last beyond the lifetime of the connection, except forthose produced by the application protocol that QUIC serves.¶

TLS session tickets and address validation tokens are used to carry QUICconfiguration information between connections, specifically, to enable a serverto efficiently recover state that is used in connection establishment andaddress validation. TheseMUST NOT be used to communicate application semanticsbetween endpoints; clientsMUST treat them as opaque values. The potential forreuse of these tokens means that they require stronger protections againstreplay.¶

A server that accepts 0-RTT on a connection incurs a higher cost than acceptinga connection without 0-RTT. This includes higher processing and computationcosts. Servers need to consider the probability of replay and all associatedcosts when accepting 0-RTT.¶

Ultimately, the responsibility for managing the risks of replay attacks with0-RTT lies with an application protocol. An application protocol that uses QUICMUST describe how the protocol uses 0-RTT and the measures that are employed toprotect against replay attack. An analysis of replay risk needs to considerall QUIC protocol features that carry application semantics.¶

Disabling 0-RTT entirely is the most effective defense against replay attack.¶

QUIC extensionsMUST either describe how replay attacks affect their operationor prohibit the use of the extension in 0-RTT. Application protocolsMUSTeither prohibit the use of extensions that carry application semantics in 0-RTTor provide replay mitigation strategies.¶

9.3.Packet Reflection Attack Mitigation

A small ClientHello that results in a large block of handshake messages from aserver can be used in packet reflection attacks to amplify the traffic generatedby an attacker.¶

QUIC includes three defenses against this attack. First, the packet containinga ClientHelloMUST be padded to a minimum size. Second, if responding to anunverified source address, the server is forbidden to send more than threetimes as many bytes as the number of bytes it has received (seeSection 8.1 of [QUIC-TRANSPORT]). Finally, because acknowledgments of Handshake packets areauthenticated, a blind attacker cannot forge them. Put together, these defenseslimit the level of amplification.¶

9.4.Header Protection Analysis

[NAN] analyzes authenticated encryptionalgorithms that provide nonce privacy, referred to as "Hide Nonce" (HN)transforms. The general header protection construction in this document isone of those algorithms (HN1). Header protection is applied after the packetprotection AEAD, sampling a set of bytes (sample) from the AEAD output andencrypting the header field using a pseudorandom function (PRF) as follows:¶

protected_field = field XOR PRF(hp_key, sample)

The header protection variants in this document use a pseudorandom permutation(PRP) in place of a generic PRF. However, since all PRPs are also PRFs[IMC],these variants do not deviate from the HN1 construction.¶

Ashp_key is distinct from the packet protection key, it follows that headerprotection achieves AE2 security as defined in[NAN] and therefore guaranteesprivacy offield, the protected packet header. Future header protectionvariants based on this constructionMUST use a PRF to ensure equivalentsecurity guarantees.¶

Use of the same key and ciphertext sample more than once risks compromisingheader protection. Protecting two different headers with the same key andciphertext sample reveals the exclusive OR of the protected fields. Assumingthat the AEAD acts as a PRF, if L bits are sampled, the odds of two ciphertextsamples being identical approach 2^-L/2, that is, the birthday bound.For the algorithms described in this document, that probability is one in2⁶⁴.¶

To prevent an attacker from modifying packet headers, the header is transitivelyauthenticated using packet protection; the entire packet header is part of theauthenticated additional data. Protected fields that are falsified or modifiedcan only be detected once the packet protection is removed.¶

9.5.Header Protection Timing Side Channels

An attacker could guess values for packet numbers or Key Phase and have anendpoint confirm guesses through timing side channels. Similarly, guesses forthe packet number length can be tried and exposed. If the recipient of a packetdiscards packets with duplicate packet numbers without attempting to removepacket protection, they could reveal through timing side channels that thepacket number matches a received packet. For authentication to be free fromside channels, the entire process of header protection removal, packet numberrecovery, and packet protection removalMUST be applied together without timingand other side channels.¶

For the sending of packets, construction and protection of packet payloads andpacket numbersMUST be free from side channels that would reveal the packetnumber or its encoded size.¶

During a key update, the time taken to generate new keys could reveal throughtiming side channels that a key update has occurred. Alternatively, where anattacker injects packets, this side channel could reveal the value of the KeyPhase on injected packets. After receiving a key update, an endpointSHOULDgenerate and save the next set of receive packet protection keys, as describedinSection 6.3. By generating new keys before a key update isreceived, receipt of packets will not create timing signals that leak the valueof the Key Phase.¶

This depends on not doing this key generation during packet processing, and itcan require that endpoints maintain three sets of packet protection keys forreceiving: for the previous key phase, for the current key phase, and for thenext key phase. Endpoints can instead choose to defer generation of the nextreceive packet protection keys until they discard old keys so that only two setsof receive keys need to be retained at any point in time.¶

9.6.Key Diversity

In using TLS, the central key schedule of TLS is used. As a result of the TLShandshake messages being integrated into the calculation of secrets, theinclusion of the QUIC transport parameters extension ensures that the handshakeand 1-RTT keys are not the same as those that might be produced by a serverrunning TLS over TCP. To avoid the possibility of cross-protocol keysynchronization, additional measures are provided to improve key separation.¶

The QUIC packet protection keys and IVs are derived using a different label thanthe equivalent keys in TLS.¶

To preserve this separation, a new version of QUICSHOULD define new labels forkey derivation for packet protection key and IV, plus the header protectionkeys. This version of QUIC uses the string "quic". Other versions can use aversion-specific label in place of that string.¶

The initial secrets use a key that is specific to the negotiated QUIC version.New QUIC versionsSHOULD define a new salt value used in calculating initialsecrets.¶

9.7.Randomness

QUIC depends on endpoints being able to generate secure random numbers, bothdirectly for protocol values such as the connection ID, and transitively viaTLS. See[RFC4086] for guidance on secure random number generation.¶

A.1.Keys

The labels generated during the execution of the HKDF-Expand-Label function(that is, HkdfLabel.label) and part of the value given to the HKDF-Expandfunction in order to produce its output are:¶

client in:

00200f746c73313320636c69656e7420696e00¶

server in:

00200f746c7331332073657276657220696e00¶

quic key:

00100e746c7331332071756963206b657900¶

quic iv:

000c0d746c733133207175696320697600¶

quic hp:

00100d746c733133207175696320687000¶

The initial secret is common:¶

initial_secret = HKDF-Extract(initial_salt, cid)    = 7db5df06e7a69e432496adedb0085192      3595221596ae2ae9fb8115c1e9ed0a44

The secrets for protecting client packets are:¶

client_initial_secret    = HKDF-Expand-Label(initial_secret, "client in", "", 32)    = c00cf151ca5be075ed0ebfb5c80323c4      2d6b7db67881289af4008f1f6c357aeakey = HKDF-Expand-Label(client_initial_secret, "quic key", "", 16)    = 1f369613dd76d5467730efcbe3b1a22div  = HKDF-Expand-Label(client_initial_secret, "quic iv", "", 12)    = fa044b2f42a3fd3b46fb255chp  = HKDF-Expand-Label(client_initial_secret, "quic hp", "", 16)    = 9f50449e04a0e810283a1e9933adedd2

The secrets for protecting server packets are:¶

server_initial_secret    = HKDF-Expand-Label(initial_secret, "server in", "", 32)    = 3c199828fd139efd216c155ad844cc81      fb82fa8d7446fa7d78be803acdda951bkey = HKDF-Expand-Label(server_initial_secret, "quic key", "", 16)    = cf3a5331653c364c88f0f379b6067e37iv  = HKDF-Expand-Label(server_initial_secret, "quic iv", "", 12)    = 0ac1493ca1905853b0bba03ehp  = HKDF-Expand-Label(server_initial_secret, "quic hp", "", 16)    = c206b8d9b9f0f37644430b490eeaa314

A.2.Client Initial

The client sends an Initial packet. The unprotected payload of this packetcontains the following CRYPTO frame, plus enough PADDING frames to make a1162-byte payload:¶

060040f1010000ed0303ebf8fa56f129 39b9584a3896472ec40bb863cfd3e86804fe3a47f06a2b69484c000004130113 02010000c000000010000e00000b6578616d706c652e636f6dff01000100000a 00080006001d0017001800100007000504616c706e0005000501000000000033 00260024001d00209370b2c9caa47fbabaf4559fedba753de171fa71f50f1ce1 5d43e994ec74d748002b0003020304000d0010000e0403050306030203080408 050806002d00020101001c00024001003900320408ffffffffffffffff050480 00ffff07048000ffff0801100104800075300901100f088394c8f03e51570806 048000ffff

The unprotected header indicates a length of 1182 bytes: the 4-byte packetnumber, 1162 bytes of frames, and the 16-byte authentication tag. The headerincludes the connection ID and a packet number of 2:¶

c300000001088394c8f03e5157080000449e00000002

Protecting the payload produces output that is sampled for header protection.Because the header uses a 4-byte packet number encoding, the first 16 bytes ofthe protected payload is sampled and then applied to the header as follows:¶

sample = d1b1c98dd7689fb8ec11d242b123dc9bmask = AES-ECB(hp, sample)[0..4]     = 437b9aec36header[0] ^= mask[0] & 0x0f     = c0header[18..21] ^= mask[1..4]     = 7b9aec34header = c000000001088394c8f03e5157080000449e7b9aec34

The resulting protected packet is:¶

c000000001088394c8f03e5157080000 449e7b9aec34d1b1c98dd7689fb8ec11d242b123dc9bd8bab936b47d92ec356c 0bab7df5976d27cd449f63300099f3991c260ec4c60d17b31f8429157bb35a12 82a643a8d2262cad67500cadb8e7378c8eb7539ec4d4905fed1bee1fc8aafba1 7c750e2c7ace01e6005f80fcb7df621230c83711b39343fa028cea7f7fb5ff89 eac2308249a02252155e2347b63d58c5457afd84d05dfffdb20392844ae81215 4682e9cf012f9021a6f0be17ddd0c2084dce25ff9b06cde535d0f920a2db1bf3 62c23e596d11a4f5a6cf3948838a3aec4e15daf8500a6ef69ec4e3feb6b1d98e 610ac8b7ec3faf6ad760b7bad1db4ba3485e8a94dc250ae3fdb41ed15fb6a8e5 eba0fc3dd60bc8e30c5c4287e53805db059ae0648db2f64264ed5e39be2e20d8 2df566da8dd5998ccabdae053060ae6c7b4378e846d29f37ed7b4ea9ec5d82e7 961b7f25a9323851f681d582363aa5f89937f5a67258bf63ad6f1a0b1d96dbd4 faddfcefc5266ba6611722395c906556be52afe3f565636ad1b17d508b73d874 3eeb524be22b3dcbc2c7468d54119c7468449a13d8e3b95811a198f3491de3e7 fe942b330407abf82a4ed7c1b311663ac69890f4157015853d91e923037c227a 33cdd5ec281ca3f79c44546b9d90ca00f064c99e3dd97911d39fe9c5d0b23a22 9a234cb36186c4819e8b9c5927726632291d6a418211cc2962e20fe47feb3edf 330f2c603a9d48c0fcb5699dbfe5896425c5bac4aee82e57a85aaf4e2513e4f0 5796b07ba2ee47d80506f8d2c25e50fd14de71e6c418559302f939b0e1abd576 f279c4b2e0feb85c1f28ff18f58891ffef132eef2fa09346aee33c28eb130ff2 8f5b766953334113211996d20011a198e3fc433f9f2541010ae17c1bf202580f 6047472fb36857fe843b19f5984009ddc324044e847a4f4a0ab34f719595de37 252d6235365e9b84392b061085349d73203a4a13e96f5432ec0fd4a1ee65accd d5e3904df54c1da510b0ff20dcc0c77fcb2c0e0eb605cb0504db87632cf3d8b4 dae6e705769d1de354270123cb11450efc60ac47683d7b8d0f811365565fd98c 4c8eb936bcab8d069fc33bd801b03adea2e1fbc5aa463d08ca19896d2bf59a07 1b851e6c239052172f296bfb5e72404790a2181014f3b94a4e97d117b4381303 68cc39dbb2d198065ae3986547926cd2162f40a29f0c3c8745c0f50fba3852e5 66d44575c29d39a03f0cda721984b6f440591f355e12d439ff150aab7613499d bd49adabc8676eef023b15b65bfc5ca06948109f23f350db82123535eb8a7433 bdabcb909271a6ecbcb58b936a88cd4e8f2e6ff5800175f113253d8fa9ca8885 c2f552e657dc603f252e1a8e308f76f0be79e2fb8f5d5fbbe2e30ecadd220723 c8c0aea8078cdfcb3868263ff8f0940054da48781893a7e49ad5aff4af300cd8 04a6b6279ab3ff3afb64491c85194aab760d58a606654f9f4400e8b38591356f bf6425aca26dc85244259ff2b19c41b9f96f3ca9ec1dde434da7d2d392b905dd f3d1f9af93d1af5950bd493f5aa731b4056df31bd267b6b90a079831aaf579be 0a39013137aac6d404f518cfd46840647e78bfe706ca4cf5e9c5453e9f7cfd2b 8b4c8d169a44e55c88d4a9a7f9474241e221af44860018ab0856972e194cd934

A.3.Server Initial

The server sends the following payload in response, including an ACK frame, aCRYPTO frame, and no PADDING frames:¶

02000000000600405a020000560303ee fce7f7b37ba1d1632e96677825ddf73988cfc79825df566dc5430b9a045a1200 130100002e00330024001d00209d3c940d89690b84d08a60993c144eca684d10 81287c834d5311bcf32bb9da1a002b00020304

The header from the server includes a new connection ID and a 2-byte packetnumber encoding for a packet number of 1:¶

c1000000010008f067a5502a4262b50040750001

As a result, after protection, the header protection sample is taken startingfrom the third protected byte:¶

sample = 2cd0991cd25b0aac406a5816b6394100mask   = 2ec0d8356aheader = cf000000010008f067a5502a4262b5004075c0d9

The final protected packet is then:¶

cf000000010008f067a5502a4262b500 4075c0d95a482cd0991cd25b0aac406a5816b6394100f37a1c69797554780bb3 8cc5a99f5ede4cf73c3ec2493a1839b3dbcba3f6ea46c5b7684df3548e7ddeb9 c3bf9c73cc3f3bded74b562bfb19fb84022f8ef4cdd93795d77d06edbb7aaf2f 58891850abbdca3d20398c276456cbc42158407dd074ee

A.4.Retry

This shows a Retry packet that might be sent in response to the Initial packetinAppendix A.2. The integrity check includes the client-chosenconnection ID value of 0x8394c8f03e515708, but that value is notincluded in the final Retry packet:¶

ff000000010008f067a5502a4262b574 6f6b656e04a265ba2eff4d829058fb3f0f2496ba

A.5.ChaCha20-Poly1305 Short Header Packet

This example shows some of the steps required to protect a packet witha short header. This example uses AEAD_CHACHA20_POLY1305.¶

In this example, TLS produces an application write secret from which a serveruses HKDF-Expand-Label to produce four values: a key, an IV, a headerprotection key, and the secret that will be used after keys are updated (thislast value is not used further in this example).¶

secret    = 9ac312a7f877468ebe69422748ad00a1      5443f18203a07d6060f688f30f21632bkey = HKDF-Expand-Label(secret, "quic key", "", 32)    = c6d98ff3441c3fe1b2182094f69caa2e      d4b716b65488960a7a984979fb23e1c8iv  = HKDF-Expand-Label(secret, "quic iv", "", 12)    = e0459b3474bdd0e44a41c144hp  = HKDF-Expand-Label(secret, "quic hp", "", 32)    = 25a282b9e82f06f21f488917a4fc8f1b      73573685608597d0efcb076b0ab7a7a4ku  = HKDF-Expand-Label(secret, "quic ku", "", 32)    = 1223504755036d556342ee9361d25342      1a826c9ecdf3c7148684b36b714881f9

The following shows the steps involved in protecting a minimal packet with anempty Destination Connection ID. This packet contains a single PING frame (thatis, a payload of just 0x01) and has a packet number of 654360564. In thisexample, using a packet number of length 3 (that is, 49140 is encoded) avoidshaving to pad the payload of the packet; PADDING frames would be needed if thepacket number is encoded on fewer bytes.¶

pn                 = 654360564 (decimal)nonce              = e0459b3474bdd0e46d417eb0unprotected header = 4200bff4payload plaintext  = 01payload ciphertext = 655e5cd55c41f69080575d7999c25a5bfb

The resulting ciphertext is the minimum size possible. One byte is skipped toproduce the sample for header protection.¶

sample = 5e5cd55c41f69080575d7999c25a5bfbmask   = aefefe7d03header = 4cfe4189

The protected packet is the smallest possible packet size of 21 bytes.¶

packet = 4cfe4189655e5cd55c41f69080575d7999c25a5bfb

B.1.Analysis of AEAD_AES_128_GCM and AEAD_AES_256_GCM Usage Limits

[GCM-MU] specifies concrete bounds for AEAD_AES_128_GCM and AEAD_AES_256_GCMas used in TLS 1.3 and QUIC. This section documents this analysis using severalsimplifying assumptions:¶

• The number of ciphertext blocks an attacker uses in forgery attempts isbounded by v * l, which is the number of forgery attempts multiplied by thesize of each packet (in blocks).¶
• The amount of offline work done by an attacker does not dominate other factorsin the analysis.¶

The bounds in[GCM-MU] are tighter and more complete than those used in[AEBounds], which allows for larger limits than those described in[TLS13].¶

2 * (q * l)^2 / 2^n

q <= 2^35 / l

(1 / 2^(8 * n)) + ((2 * v) / 2^(2 * n))        + ((2 * o * v) / 2^(k + n)) + (n * (v + (v * l)) / 2^k)

v <= 2^64 / l

v <= 2^192 / l

B.2.Analysis of AEAD_AES_128_CCM Usage Limits

TLS[TLS13] and[AEBounds] do not specify limits on usagefor AEAD_AES_128_CCM. However, any AEAD that is used with QUIC requires limitson use that ensure that both confidentiality and integrity are preserved. Thissection documents that analysis.¶

[CCM-ANALYSIS] is used as the basis of thisanalysis. The results of that analysis are used to derive usage limits that arebased on those chosen in[TLS13].¶

For confidentiality, Theorem 2 in[CCM-ANALYSIS] establishes that an attackergains a distinguishing advantage over an ideal pseudorandom permutation (PRP) ofno more than the following:¶

(2l * q)^2 / 2^n

The integrity limit in Theorem 1 in[CCM-ANALYSIS] provides an attacker astrictly higher advantage for the same number of messages. As the targets forthe confidentiality advantage and the integrity advantage are the same, onlyTheorem 1 needs to be considered.¶

Theorem 1 establishes that an attacker gains an advantage over anideal PRP of no more than the following:¶

v / 2^t + (2l * (v + q))^2 / 2^n

Ast andn are both 128, the first term is negligible relative to thesecond, so that term can be removed without a significant effect on the result.¶

This produces a relation that combines both encryption and decryption attemptswith the same limit as that produced by the theorem for confidentiality alone.For a target advantage of 2^-57, this results in the following:¶

v + q <= 2^34.5 / l

By settingq = v, values for both confidentiality and integrity limits can beproduced. Endpoints that limit packets to 2¹¹ bytes therefore haveboth confidentiality and integrity limits of 2^26.5 packets. Endpointsthat do not restrict packet size have a limit of 2^21.5.¶