RFC 9605
Secure Frame (SFrame): Lightweight Authenticated Encryption for Real-Time Media
Part 1 of 11 🎀
RFC 9605: Secure Frame (SFrame) 💖
Internet Engineering Task Force (IETF) 🌸 Request for Comments: 9605 ✨ Category: Standards Track (so it's an official Internet standard, bestie) 👑 ISSN: 2070-1721 Date: August 2024 📅 Obsoletes / Updates: The title block doesn't list anything it obsoletes or updates 🫶
Authors: E. Omara (Apple), J. Uberti (Fixie.ai), S. G. Murillo (CoSMo Software), R. Barnes, Ed. (Cisco), and Y. Fablet (Apple) 🧚
Title: Secure Frame (SFrame): Lightweight Authenticated Encryption for Real-Time Media 🦋
Abstract 🌈
okay so this doc describes SFrame, an end-to-end encryption and authentication mechanism for media frames in multiparty conference calls 📞. The vibe is that central media servers (Selective Forwarding Units, or SFUs) can read the media metadata they need to make forwarding decisions, but they still can't see the actual media 🙈.
SFrame is not the same as the Secure Real-Time Protocol (SRTP) 💅. It's independent of RTP, which means it works with non-RTP media transport too, and it can be applied to whole media frames so it's more bandwidth efficient 🔥.
Status of This Memo ✨
This is an Internet Standards Track document 👑.
This document comes from the Internet Engineering Task Force (IETF) 🌸. It represents the consensus of the IETF community, got public review, and was approved for publication by the Internet Engineering Steering Group (IESG) 🫶. If you want more info on Internet Standards, check Section 2 of RFC 7841 💜.
For the current status of this document, any errata, and how to give feedback on it, go to https://www.rfc-editor.org/info/rfc9605 🦋.
Copyright Notice 🎀
Copyright (c) 2024 IETF Trust and the persons identified as the document authors. All rights reserved 💖.
This document is subject to BCP 78 and the IETF Trust's Legal Provisions Relating to IETF Documents (https://trustee.ietf.org/license-info) as they stand on the date of publication 📜. Please actually read these docs, bestie, because they spell out your rights and restrictions for this document 🙈. Code Components pulled out of this document must include the Revised BSD License text described in Section 4.e of the Trust Legal Provisions, and they come with no warranty, as the Revised BSD License describes 💅.
Table of Contents 🧚
1. Introduction
2. Terminology
3. Goals
4. SFrame
4.1. Application Context
4.2. SFrame Ciphertext
4.3. SFrame Header
4.4. Encryption Schema
4.4.1. Key Selection
4.4.2. Key Derivation
4.4.3. Encryption
4.4.4. Decryption
4.5. Cipher Suites
4.5.1. AES-CTR with SHA2
5. Key Management
5.1. Sender Keys
5.2. MLS
6. Media Considerations
6.1. Selective Forwarding Units
6.1.1. RTP Stream Reuse
6.1.2. Simulcast
6.1.3. Scalable Video Coding (SVC)
6.2. Video Key Frames
6.3. Partial Decoding
7. Security Considerations
7.1. No Header Confidentiality
7.2. No Per-Sender Authentication
7.3. Key Management
7.4. Replay
7.5. Risks Due to Short Tags
8. IANA Considerations
8.1. SFrame Cipher Suites
9. Application Responsibilities
9.1. Header Value Uniqueness
9.2. Key Management Framework
9.3. Anti-Replay
9.4. Metadata
10. References
10.1. Normative References
10.2. Informative References
Appendix A. Example API
Appendix B. Overhead Analysis
B.1. Assumptions
B.2. Audio
B.3. Video
B.4. Conferences
B.5. SFrame over RTP
Appendix C. Test Vectors
C.1. Header Encoding/Decoding
C.2. AEAD Encryption/Decryption Using AES-CTR and HMAC
C.3. SFrame Encryption/Decryption
Acknowledgements
Contributors
Authors' Addresses1. Introduction ✨
Modern multiparty video call systems use Selective Forwarding Unit (SFU) servers to route media streams to call endpoints efficiently 📞. The SFU picks where things go based on stuff like available bandwidth, desired video size, codec support, and other factors 🧠. An SFU typically doesn't need to access the conference's media content, so the media can be encrypted "end to end" and the SFU can't decrypt it 🔒.
Here's the catch, though 💀. For the SFU to work properly, it usually needs to access RTP metadata and RTCP feedback messages. That's not possible if all RTP/RTCP traffic is end-to-end encrypted 🙈.
So two layers of encryption and authentication are required 🫶:
- Hop-by-hop (HBH) encryption of media, metadata, and feedback messages between the endpoints and SFU 🦋
- End-to-end (E2E) encryption (E2EE) of media between the endpoints 💜
The Secure Real-Time Protocol (SRTP) is already super popular for HBH encryption [RFC3711] 🌸. The SRTP "double encryption" scheme is a way to do E2E encryption in SRTP [RFC8723]. Unfortunately, it's kinda inefficient and really complicated, and because it's so tangled up with RTP, it just doesn't work in several realistic SFU scenarios 😭.
So this document proposes a brand new E2EE protection scheme called SFrame 🎀. It's specifically designed for group conference calls with SFUs. SFrame is a general encryption framing that can protect media payloads, and it doesn't care what transport you're using 🧚.
2. Terminology 💅
The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "NOT RECOMMENDED", "MAY", and "OPTIONAL" are the BCP 14 words, and they mean what BCP 14 [RFC2119] [RFC8174] says they mean, but only when they show up in ALL CAPS like this 📣. Lowercase versions are just regular words 🙈.
Quick acronym check 🍓:
- MAC: Message Authentication Code
- E2EE: End-to-End Encryption
- HBH: Hop-by-Hop
We use "Selective Forwarding Unit (SFU)" and "media stream" in a looser way than [RFC7656] does 🦋. An SFU is a selective switching function for media payloads. A media stream is a sequence of media payloads, whether those payloads ride over RTP or some other protocol ✨.
3. Goals 🎯
SFrame is meant to be a good E2EE protection scheme for conference call media in a wide range of scenarios. Its goals are:
- Provide a secure E2EE mechanism for audio and video in conference calls that can be used with arbitrary SFU servers 🔐.
- Decouple media encryption from key management, so SFrame can be used with an arbitrary key management system 🗝️.
- Minimize packet expansion, so conferencing can work in as many network conditions as possible 📶.
- Decouple the media encryption framework from the underlying transport, so it can be used in non-RTP scenarios, e.g., WebTransport [WEBTRANSPORT] 🌐.
- When used with RTP and its error-resilience mechanisms, i.e., RTX and Forward Error Correction (FEC), require no special handling for RTX and FEC packets 🛡️.
- Minimize the changes needed in SFU servers 🖥️.
- Minimize the changes needed in endpoints 📱.
- Work with the most popular audio and video codecs used in conferencing scenarios 🎬.
4. SFrame 💜
This document defines an encryption mechanism that gives you effective E2EE 🔒. It's simple to implement, has no dependencies on RTP, and keeps the encryption bandwidth overhead low 📉. This section explains how the mechanism works, including how applications use SFrame to protect media and the actual mechanics of E2EE for protecting media 🧚.
The text you pasted is the RFC body itself, not a shell or file task, so there's nothing to run here. Continuing the rewrite from Section 4.4, here's part 2 in the bestie voice. 🎀
4.1. Application Context ✨
okay so SFrame is a general encryption framing that's basically built to be an E2EE layer sitting on top of a transport that already does HBH encryption, like SRTP or QUIC [RFC3711][MOQ-TRANSPORT]. the vibe is layered protection, bestie 🧚
how much SFrame costs you depends on the scale you encrypt at. there are two common patterns: encrypting whole media frames (per frame) or encrypting the individual transport-level media payloads (per packet). the overhead and how hard it is to integrate both depend on which one you pick 💅
here's a typical media sender stack: it grabs media from a source, encodes it into frames, splits those frames into media packets, and then sends the payloads inside SRTP packets. the receiver just does everything in reverse, putting frames back together from SRTP packets and then decoding. Figure 1 shows the two spots where SFrame can slot in, either around whole frames or around individual media packets 🦋
per frame is the more efficient option, but it can get messy to integrate in environments where depacketization depends on what's inside the media packets. per packet skips that headache, but you pay for it in higher bandwidth. there's some quantitative tea on these trade-offs in Appendix B 🍓
also, SFrame is a general media encapsulation, so it works in other scenarios too. the main thing is that the sender and receivers of an SFrame-encrypted object have to agree on what that object means. SFrame does NOT handle that agreement for you, the application has to set it up 🫶
+------------------------------------------------------+
| |
| +--------+ +-------------+ +-----------+ |
.-. | | | | | | HBH | |
| | | | Encode |----->| Packetize |----->| Protect |----------+
'+' | | | ^ | | ^ | | | |
/|\ | +--------+ | +-------------+ | +-----------+ | |
/ + \ | | | ^ | |
/ \ | SFrame SFrame | | |
/ \ | Protect Protect | | |
Alice | (per frame) (per packet) | | |
| ^ ^ | | |
| | | | | |
+---------------|-------------------|---------|--------+ |
| | | v
| | | +------+-+
| E2E Key | HBH Key | Media |
+---- Management ---+ Management | Server |
| | | +------+-+
| | | |
+---------------|-------------------|---------|--------+ |
| | | | | |
| V V | | |
.-. | SFrame SFrame | | |
| | | Unprotect Unprotect | | |
'+' | (per frame) (per packet) | | |
/|\ | | | V | |
/ + \ | +--------+ | +-------------+ | +-----------+ | |
/ \ | | | V | | V | HBH | | |
/ \ | | Decode |<-----| Depacketize |<-----| Unprotect |<---------+
Bob | | | | | | | |
| +--------+ +-------------+ +-----------+ |
| |
+------------------------------------------------------+Figure 1: Two Options for Integrating SFrame in a Typical Media Stack
ngl this is the part where it gets real: like SRTP, SFrame does NOT define how the parties in the conference exchange their SFrame keys. keys might come over an existing E2E-secure channel (see Section 5.1), or get derived from an E2E-secure shared secret (see Section 5.2). the key management system MUST ensure that each key used for encrypting media is used by exactly one media sender, so nonces never get reused. 💀🔥
4.2. SFrame Ciphertext 🌈
an SFrame ciphertext is an SFrame header followed by the output of an Authenticated Encryption with Associated Data (AEAD) encryption of the plaintext [RFC5116], with the header given as additional authenticated data (AAD). the header is a variable-length structure, and you can read all about it in Section 4.3 💜 the layout of the encrypted data and the authentication tag is decided by whichever AEAD algorithm you're using.
+-+----+-+----+--------------------+--------------------+<-+
|K|KLEN|C|CLEN| Key ID | Counter | |
+->+-+----+-+----+--------------------+--------------------+ |
| | | |
| | | |
| | | |
| | | |
| | Encrypted Data | |
| | | |
| | | |
| | | |
| | | |
+->+-------------------------------------------------------+<-+
| | Authentication Tag | |
| +-------------------------------------------------------+ |
| |
| |
+--- Encrypted Portion Authenticated Portion ---+Figure 2: Structure of an SFrame Ciphertext
when SFrame is applied per packet, each packet's payload is an SFrame ciphertext. when it's applied per frame, the SFrame ciphertext for one encrypted frame spans several packets: the header shows up in the first packet and the authentication tag shows up in the last one. reassembling that frame from the individual packets, dealing with packet loss and reordering as needed, is the application's job 🎀
4.3. SFrame Header 💖
the SFrame header carries two values that the encryption parameters get derived from:
- A Key ID (KID) that determines which encryption key should be used
- A Counter (CTR) that is used to construct the nonce for the encryption
applications MUST make sure each (KID, CTR) combination is used for exactly one SFrame encryption operation. Section 9.1 walks through a typical way to get that guarantee 🔐
Config Byte
|
.-----' '-----.
| |
0 1 2 3 4 5 6 7
+-+-+-+-+-+-+-+-+------------+------------+
|X| K |Y| C | KID... | CTR... |
+-+-+-+-+-+-+-+-+------------+------------+
Figure 3: SFrame Headerthe SFrame header has the overall structure in Figure 3. the first byte is a "config byte" with these fields:
- Extended KID Flag (X, 1 bit): tells you whether the K field holds the KID or the KID length.
- KID or KID Length (K, 3 bits): if X is 0, this field holds the KID. if X is 1, it holds the length of the KID, minus one.
- Extended CTR Flag (Y, 1 bit): tells you whether the C field holds the CTR or the CTR length.
- CTR or CTR Length (C, 3 bits): this field holds the CTR if Y is 0, or the CTR length, minus one, if Y is 1.
the KID and CTR fields are compact unsigned integers in network (big-endian) byte order. if a value is in the range 0-7, it rides in the matching bits of the config byte (K or C), and the matching flag (X or Y) is set to zero. otherwise, the value MUST be encoded with the minimum number of bytes required and tacked on after the config byte, KID first and CTR second. the header field (K or C) is set to the number of bytes in the encoded value, minus one. so 000 means a length of 1, 001 means a length of 2, and so on. that means a 3-bit length field can cover value lengths 1-8 📏✨
the SFrame header can take one of the four forms in Figure 4, depending on which X and Y flags are set.
KID < 8, CTR < 8:
+-+-----+-+-----+
|0| KID |0| CTR |
+-+-----+-+-----+
KID < 8, CTR >= 8:
+-+-----+-+-----+------------------------+
|0| KID |1|CLEN | CTR... (length=CLEN) |
+-+-----+-+-----+------------------------+
KID >= 8, CTR < 8:
+-+-----+-+-----+------------------------+
|1|KLEN |0| CTR | KID... (length=KLEN) |
+-+-----+-+-----+------------------------+
KID >= 8, CTR >= 8:
+-+-----+-+-----+------------------------+------------------------+
|1|KLEN |1|CLEN | KID... (length=KLEN) | CTR... (length=CLEN) |
+-+-----+-+-----+------------------------+------------------------+
Figure 4: Forms of Encoded SFrame Header4.4. Encryption Schema 🧁
SFrame encryption uses an AEAD encryption algorithm and a hash function, both defined by the cipher suite in use (see Section 4.5). here are the bits of the AEAD and hash algorithm we'll keep referring to:
- AEAD.Encrypt and AEAD.Decrypt: the encryption and decryption functions for the AEAD. we follow the convention of RFC 5116 [RFC5116] and treat the authentication tag as part of the ciphertext produced by AEAD.Encrypt (unlike SRTP [RFC3711], which keeps it as a separate field).
- AEAD.Nk: the size in bytes of a key for the encryption algorithm
- AEAD.Nn: the size in bytes of a nonce for the encryption algorithm
- AEAD.Nt: the overhead in bytes of the encryption algorithm (usually the size of the "tag" added to the plaintext)
- AEAD.Nka: for cipher suites using the compound AEAD described in Section 4.5.1, the size in bytes of a key for the underlying encryption algorithm
- Hash.Nh: the size in bytes of the output of the hash function
4.4.1. Key Selection ✨
Okay so every SFrame encryption or decryption op is premised on one secret base_key 🔑, and each one is labeled with an integer KID value that gets signaled in the SFrame header. 🎀
The sender and receivers need to agree on which base_key goes with which KID 🤝. They also need to agree on whether a base_key is for encryption or decryption only. How you provision base_key values and their KID values is beyond the scope of this specification, bestie, but its security properties will bound the assurances SFrame gives you 😭. For example, if SFrame is used to provide E2E security against intermediary media nodes, then SFrame keys need to be negotiated in a way that does not make them accessible to those intermediaries 🙈.
For each known KID value, the client stores the matching symmetric key base_key. For keys that can be used for encryption, the client also stores the next CTR value to be used when encrypting (initially 0) 🍓.
When encrypting, the application specifies which KID to use, and the CTR value goes up after successful encryption. When decrypting, the base_key is selected from the available keys using the KID value in the SFrame header 🧚.
A given base_key MUST NOT be used for encryption by multiple senders. Reuse would mean multiple encrypted frames get generated with the same (key, nonce) pair, and that harms the protections provided by many AEAD algorithms 💀. Implementations MUST mark each base_key as usable for encryption or decryption, never both 💅.
Note that the set of available keys might change over the lifetime of a real-time session. In those cases, the client will need to manage key usage so it doesn't lose media, because a key being used to encrypt before all receivers can use it to decrypt is a recipe for drama 😩. For example, an application may make decryption-only keys available immediately, but delay the use of keys for encryption until (a) all receivers have acknowledged receipt of the new key, or (b) a timeout expires ⏳.
4.4.2. Key Derivation 🧪
SFrame encryption and decryption use a key and salt derived from the base_key associated with a KID. Given a base_key value, the key and salt are derived using HMAC-based Key Derivation Function (HKDF) [RFC5869] as follows:
def derive_key_salt(KID, base_key):
sframe_secret = HKDF-Extract("", base_key)
sframe_key_label = "SFrame 1.0 Secret key " + KID + cipher_suite
sframe_key =
HKDF-Expand(sframe_secret, sframe_key_label, AEAD.Nk)
sframe_salt_label = "SFrame 1.0 Secret salt " + KID + cipher_suite
sframe_salt =
HKDF-Expand(sframe_secret, sframe_salt_label, AEAD.Nn)
return sframe_key, sframe_saltIn the derivation of sframe_secret:
- The + operator represents concatenation of byte strings. 🎀
- The KID value is encoded as an 8-byte big-endian integer, not the compressed form used in the SFrame header. 💜
- The cipher_suite value is a 2-byte big-endian integer representing the cipher suite in use (see Section 8.1). 🌸
The hash function used for HKDF is determined by the cipher suite in use. 🌈
4.4.3. Encryption 🔐
SFrame encryption uses the AEAD encryption algorithm for the cipher suite in use. The key for the encryption is the sframe_key. The nonce is formed by first XORing the sframe_salt with the current CTR value, and then encoding the result as a big-endian integer of length AEAD.Nn 🧁.
The encryptor forms an SFrame header using the CTR and KID values provided. The encoded header is provided as AAD to the AEAD encryption operation, together with application-provided metadata about the encrypted media (see Section 9.4) 🦋.
def encrypt(CTR, KID, metadata, plaintext):
sframe_key, sframe_salt = key_store[KID]
# encode_big_endian(x, n) produces an n-byte string encoding the
# integer x in big-endian byte order.
ctr = encode_big_endian(CTR, AEAD.Nn)
nonce = xor(sframe_salt, CTR)
# encode_sframe_header produces a byte string encoding the
# provided KID and CTR values into an SFrame header.
header = encode_sframe_header(CTR, KID)
aad = header + metadata
ciphertext = AEAD.Encrypt(sframe_key, nonce, aad, plaintext)
return header + ciphertextFor example, the metadata input to encryption lets frame metadata get authenticated when SFrame is applied per frame. After encoding the frame and before packetizing it, the necessary media metadata gets moved out of the encoded frame buffer to be sent in some channel visible to the SFU (e.g., an RTP header extension) 🧚♀️.
+---------------+
| |
| |
| plaintext |
| |
| |
+-------+-------+
|
.- +-----+ |
| | +--+--> sframe_key ----->| Key
Header | | KID | | |
| | | +--> sframe_salt --+ |
+--+ +-----+ | |
| | | +---------------------+->| Nonce
| | | CTR | |
| | | | |
| '- +-----+ |
| |
| +----------------+ |
| | metadata | |
| +-------+--------+ |
| | |
+------------------+----------------->| AAD
| |
| AEAD.Encrypt
| |
| SFrame Ciphertext |
| +---------------+ |
+-------------->| SFrame Header | |
+---------------+ |
| | |
| |<----+
| ciphertext |
| |
| |
+---------------+Figure 5: Encrypting an SFrame Ciphertext
4.4.4. Decryption 🕵️♀️
Before decrypting, a receiver needs to assemble a full SFrame ciphertext. When an SFrame ciphertext is fragmented into multiple parts for transport (e.g., a whole encrypted frame sent in multiple SRTP packets), the receiving client collects all the fragments of the ciphertext, using appropriate sequencing and start/end markers in the transport. Once all of the required fragments are available, the client reassembles them into the SFrame ciphertext and passes the ciphertext to SFrame for decryption 🧩.
The KID field in the SFrame header is used to find the right key and salt for the encrypted frame, and the CTR field is used to construct the nonce. The SFrame decryption procedure is as follows 🔓:
def decrypt(metadata, sframe_ciphertext):
KID, CTR, header, ciphertext = parse_ciphertext(sframe_ciphertext)
sframe_key, sframe_salt = key_store[KID]
ctr = encode_big_endian(CTR, AEAD.Nn)
nonce = xor(sframe_salt, ctr)
aad = header + metadata
return AEAD.Decrypt(sframe_key, nonce, aad, ciphertext)If a ciphertext fails to decrypt because there is no key available for the KID in the SFrame header, the client MAY buffer the ciphertext and retry decryption once a key with that KID is received 🪣. If a ciphertext fails to decrypt for any other reason, the client MUST discard the ciphertext 🗑️. Invalid ciphertexts SHOULD be discarded in a way that is indistinguishable (to an external observer) from having processed a valid ciphertext. In other words, the SFrame decrypt operation should take the same amount of time regardless of whether decryption succeeds or fails ⏱️.
🚨 Timing matters here, bestie. Whether decryption succeeds or fails, the operation should take the same amount of time, so nobody on the outside can tell which one happened. 🙈
SFrame Ciphertext
+---------------+
+---------------| SFrame Header |
| +---------------+
| | |
| | |-----+
| | ciphertext | |
| | | |
| | | |
| +---------------+ |
| |
| .- +-----+ |
| | | +--+--> sframe_key ----->| Key
| | | KID | | |
| | | | +--> sframe_salt --+ |
+->+ +-----+ | |
| | | +---------------------+->| Nonce
| | | CTR | |
| | | | |
| '- +-----+ |
| |
| +----------------+ |
| | metadata | |
| +-------+--------+ |
| | |
+------------------+----------------->| AAD
|
AEAD.Decrypt
|
V
+---------------+
| |
| |
| plaintext |
| |
| |
+---------------+Figure 6: Decrypting an SFrame Ciphertext
4.5. Cipher Suites ✨
Okay so cipher suites, bestie 💅. Each SFrame session uses a single cipher suite, and that suite bundles together two primitives:
- a hash function for key derivation 🔑
- an AEAD encryption algorithm [RFC5116] for frame encryption, optionally with a truncated authentication tag 🎀
This document defines the following cipher suites, with the constants defined in Section 4.4:
+============================+====+=====+====+====+====+
| Name | Nh | Nka | Nk | Nn | Nt |
+============================+====+=====+====+====+====+
| AES_128_CTR_HMAC_SHA256_80 | 32 | 16 | 48 | 12 | 10 |
+----------------------------+----+-----+----+----+----+
| AES_128_CTR_HMAC_SHA256_64 | 32 | 16 | 48 | 12 | 8 |
+----------------------------+----+-----+----+----+----+
| AES_128_CTR_HMAC_SHA256_32 | 32 | 16 | 48 | 12 | 4 |
+----------------------------+----+-----+----+----+----+
| AES_128_GCM_SHA256_128 | 32 | n/a | 16 | 12 | 16 |
+----------------------------+----+-----+----+----+----+
| AES_256_GCM_SHA512_128 | 64 | n/a | 32 | 12 | 16 |
+----------------------------+----+-----+----+----+----+
Table 1: SFrame Cipher Suite ConstantsThe numeric identifiers for these suites live in the IANA registry created in Section 8.1 🦋.
Here's a fun trick: the last number in each suite name tells you the authentication tag length. "_128" means a 128-bit tag, "_80" means 80-bit, "_64" means 64-bit, and "_32" means 32-bit 🧚.
Also, a session with multiple media streams can configure different cipher suites for different streams. For example, to conserve bandwidth, a session might use a suite with 80-bit tags for video frames and a different suite with 32-bit tags for audio frames 🎧💜.
4.5.1. AES-CTR with SHA2
Okay, this one is giving main character energy 👑. To allow very short tag sizes, we define a synthetic AEAD function that uses the authenticated counter mode of AES together with HMAC for authentication. We use an encrypt-then-MAC approach, as in SRTP [RFC3711] 🔐.
Before any encryption or decryption, the encryption and authentication subkeys get derived from the single AEAD key. The overall length of the AEAD key is Nka + Nh, where Nka is the key size for the AES block cipher in use and Nh is the output size of the hash function (as in Section 4.4). The encryption subkey is made of the first Nka bytes, and the authentication subkey is made of the remaining Nh bytes ✨.
def derive_subkeys(sframe_key):
# The encryption key comprises the first Nka bytes
enc_key = sframe_key[..Nka]
# The authentication key comprises Nh remaining bytes
auth_key = sframe_key[Nka..]
return enc_key, auth_keyThe AEAD encryption and decryption functions are then composed of individual calls to the CTR encrypt function and HMAC. The resulting MAC value gets truncated to a number of bytes Nt, which is fixed by the cipher suite 🦋.
def truncate(tag, n):
# Take the first `n` bytes of `tag`
return tag[..n]
def compute_tag(auth_key, nonce, aad, ct):
aad_len = encode_big_endian(len(aad), 8)
ct_len = encode_big_endian(len(ct), 8)
tag_len = encode_big_endian(Nt, 8)
auth_data = aad_len + ct_len + tag_len + nonce + aad + ct
tag = HMAC(auth_key, auth_data)
return truncate(tag, Nt)
def AEAD.Encrypt(key, nonce, aad, pt):
enc_key, auth_key = derive_subkeys(key)
initial_counter = nonce + 0x00000000 # append four zero bytes
ct = AES-CTR.Encrypt(enc_key, initial_counter, pt)
tag = compute_tag(auth_key, nonce, aad, ct)
return ct + tag
def AEAD.Decrypt(key, nonce, aad, ct):
inner_ct, tag = split_ct(ct, tag_len)
enc_key, auth_key = derive_subkeys(key)
candidate_tag = compute_tag(auth_key, nonce, aad, inner_ct)
if !constant_time_equal(tag, candidate_tag):
raise Exception("Authentication Failure")
initial_counter = nonce + 0x00000000 # append four zero bytes
return AES-CTR.Decrypt(enc_key, initial_counter, inner_ct)5. Key Management 🔐
Okay, so SFrame has to be integrated with an E2E key management framework, and that framework is what exchanges and rotates the keys used for SFrame encryption 🗝️. The key management framework provides these functions:
- Provisioning KID / base_key mappings to participating clients
- Updating the above data as clients join or leave
It's the application's job to provide the key management framework, as described in Section 9.2 💅.
5.1. Sender Keys
If the participants in a call already have a preexisting E2E-secure channel, they can use it to distribute SFrame keys 💌. Each client in a call generates a fresh base_key value that it will use to encrypt media. The client then uses the E2E-secure channel to send its encryption key to the other participants.
In this scheme, it's assumed that receivers have a signal outside of SFrame that says which client sent a given frame (for example, an RTP synchronization source (SSRC)). SFrame KID values are then used to tell apart versions of the sender's base_key 🎀.
KID values in this scheme have two parts: a "key generation" and a "ratchet step". Both are unsigned integers that begin at zero. The key generation goes up each time the sender distributes a new key to receivers. The ratchet step goes up each time the sender ratchets their key forward for forward secrecy 🦋:
base_key[i+1] = HKDF-Expand(
HKDF-Extract("", base_key[i]),
"SFrame 1.0 Ratchet", CipherSuite.Nh)For compactness, we don't send the whole ratchet step. Instead, we send only its low-order R bits, where R is a value set by the application. Different senders may use different values of R, but each receiver of a given sender needs to know which value of R the sender is using, so they can recognize when they need to ratchet (vs. expecting a new key) 🧠. R effectively defines a reordering window, since no more than 2^R ratchet steps can be active at a given time. The key generation is sent in the remaining 64 - R bits of the KID 🌸.
KID = (key_generation << R) + (ratchet_step % (1 << R))
64-R bits R bits
<---------------> <------------>
+-----------------+--------------+
| Key Generation | Ratchet Step |
+-----------------+--------------+
Figure 7: Structure of a KID in the Sender Keys SchemeThe sender signals a ratchet step update by sending with a KID value in which the ratchet step has been incremented. A receiver who gets a new KID from a sender computes the new key as above. The old key may be kept for some time to allow for out-of-order delivery, but should be deleted promptly 🫶.
If a new participant joins in the middle of a session, they will need to receive from each sender (a) the current sender key for that sender and (b) the current KID value for the sender 🧚. Evicting a participant requires each sender to send a fresh sender key to all receivers 💀.
It's the application's responsibility to decide when sender keys get updated 🍓. A sender key may be updated by sending a new base_key (which updates the key generation) or by hashing the current base_key (which updates the ratchet step). Ratcheting the key forward is useful when adding new receivers to an SFrame-based interaction, since it ensures that the new receivers can't decrypt any media encrypted before they were added 😭. If a sender wants the opposite property when removing a receiver (ensuring that the receiver can't decrypt media after they're removed), then the sender will need to distribute a new sender key 🔥.
5.2. MLS
The Messaging Layer Security (MLS) protocol provides group authenticated key exchange [MLS-ARCH] [MLS-PROTO] 🌈. In principle, it could be used to instantiate the sender key scheme above, but it can also be used more efficiently directly ✨.
MLS creates a linear sequence of keys, and each one is shared among the members of a group at a given point in time. When a member joins or leaves the group, a new key is produced that's known only to the augmented or reduced group. Each step in the lifetime of the group is known as an "epoch", and each member of the group is assigned an "index" that stays constant for the whole time they're in the group 🎀.
To generate keys and nonces for SFrame, we use the MLS exporter function to generate a base_key value for each MLS epoch. Each member of the group is assigned a set of KID values, so that each member has a unique sframe_key and sframe_salt to encrypt with. Senders may choose any KID value within their assigned set of KID values. For example, that lets a single sender send multiple, uncoordinated outbound media streams 💖.
base_key = MLS-Exporter("SFrame 1.0 Base Key", "", AEAD.Nk)For compactness, we don't send the whole epoch number. Instead, we send only its low-order E bits, where E is a value set by the application. E effectively defines a reordering window, since no more than 2^E epochs can be active at a given time. To handle rollover of the epoch counter, receivers MUST remove an old epoch when a new epoch with the same low-order E bits is introduced 🚨.
Let S be the number of bits required to encode a member index in the group, i.e., the smallest value such that group_size <= (1 << S). The sender index is encoded in the S bits above the epoch. The remaining 64 - S - E bits of the KID value are a context value chosen by the sender (context value 0 will produce the shortest encoded KID) 🦋.
KID = (context << (S + E)) + (sender_index << E) + (epoch % (1 << E))
64-S-E bits S bits E bits
<-----------> <------> <------>
+-------------+--------+-------+
| Context ID | Index | Epoch |
+-------------+--------+-------+
Figure 8: Structure of a KID for an MLS SenderOnce an SFrame stack has been provisioned with the sframe_epoch_secret for an epoch, it can compute the required KID values on demand (along with the resulting SFrame keys and nonces derived from the base_key and KID) whenever it needs to encrypt or decrypt for a given member 🧁.
...
|
|
Epoch 14 +--+-- index=3 ---> KID = 0x3e
| |
| +-- index=7 ---> KID = 0x7e
| |
| +-- index=20 --> KID = 0x14e
|
|
Epoch 15 +--+-- index=3 ---> KID = 0x3f
| |
| +-- index=5 ---> KID = 0x5f
|
|
Epoch 16 +----- index=2 --+--> context = 2 --> KID = 0x820
| |
| +--> context = 3 --> KID = 0xc20
|
|
Epoch 17 +--+-- index=33 --> KID = 0x211
| |
| +-- index=51 --> KID = 0x331
|
|
...
Figure 9: An Example Sequence of KIDs for an MLS-based SFrame
Session (E=4; S=6, Allowing for 64 Group Members)6. Media Considerations 🎬
6.1. Selective Forwarding Units
SFUs (e.g., those described in Section 3.7 of [RFC7667]) receive the media streams from each participant and pick which ones should be forwarded to each of the other participants 📺. There are several approaches for stream selection, but in general, the SFU needs to access metadata associated with each frame and modify the RTP information of the incoming packets when they're transmitted to the receiving participants 🧚.
This section describes how these normal SFU modes of operation interact with the E2EE provided by SFrame 💜.
6.1.1. RTP Stream Reuse ✨
okay so the SFU can totally decide to only send a certain number of streams based on who's actually talking 🎙️. to skip the overhead of setting up brand-new transport streams every time, it might just reuse streams that already exist, or even pre-allocate a set number and pick in each moment whose media goes through each one 🔄.
that means the same transport-level stream (like an RTP stream identified by either an SSRC or a Media Identification (MID)) can carry media from different streams belonging to different participants 👀. but here's the magic 🪄: every participant encrypts with their own key, so the receiver can verify who sent the media inside that RTP stream at any given moment. the receiver links the media to the sender named by the authenticated SFrame KID value, no matter how the SFU is sending stuff over to the client 💅.
ngl, there's one catch 🙈. to stop a malicious participant (not the SFU) from impersonating someone, you'd need a mechanism based on digital signatures. SFrame does not protect against that kind of attack, so don't expect it to 🚫.
6.1.2. Simulcast 🎞️
when you're doing simulcast, the same input image becomes N different encoded frames (one for each simulcast layer) 🌈. each of those frames gets processed on its own by the frame encryptor, and each one is given a unique CTR value 🔐.
6.1.3. Scalable Video Coding (SVC) 🦋
in both temporal and spatial scalability, the SFU can drop layers to match a certain bitrate, or to forward specific media sizes or frames per second 📉. so the SFU can selectively remove layers, the sender MUST encapsulate each layer in a different SFrame ciphertext 🧩. that's the non-negotiable part, bestie 🫶.
6.2. Video Key Frames 🔑
forward security and post-compromise security require that the E2EE keys (base keys) get updated any time a participant joins or leaves the call 🔄✨. this is giving "rotate on every entrance and exit" energy 💃.
the key exchange happens asynchronously, on a different path from the SFU signaling and media 📬. so it's possible that when a new participant joins and the SFU asks for a key frame, the sender ends up generating the E2EE frame with a key the receiver doesn't know yet, and the frame gets discarded 😭. when the sender updates its sending key to the new one, it sends that in a non-key frame. the receiver can decrypt it, but it can't decode it 🙃.
the new receiver then re-requests a key frame, but thanks to sender and SFU policies, that new key frame could take a while to show up ⏳.
if the sender sends a key frame after the new E2EE key is in use, the new participant sees video way faster 🏃♀️💨.
this doesn't happen for media streams without dependencies among frames, like audio streams 🎧. in those, every frame can be decoded independently, so a frame never depends on another frame sitting on the other side of a key rotation. audio is literally unbothered 💅.
6.3. Partial Decoding 🧩
some codecs support partial decoding, where individual packets can be decoded without waiting for the whole frame to arrive 📦. when SFrame is applied per frame, though, partial decoding isn't possible, because the decoder can't touch any data until the entire frame has arrived and been decrypted 🙅♀️.
7. Security Considerations 🛡️
okay bestie, this section is seriously important, so let's get into it 👀💜.
7.1. No Header Confidentiality 🙈
SFrame gives integrity protection to the SFrame header (the KID and CTR values), but it does NOT give confidentiality protection 😬. anyone who can see the SFrame header can learn things like which parties are sending SFrame payloads (from the KID values) and at what rates (from the CTR values) 📊. if SFrame is used for end-to-end security on top of hop-by-hop protections (like running over SRTP, as described in Appendix B.5), the hop-by-hop security mechanisms give confidentiality protection for the SFrame header between hops 🔒.
7.2. No Per-Sender Authentication 🚨
SFrame does not provide per-sender authentication of media data 😱. any sender in a session can send media that gets associated with any other sender. the reason is that SFrame uses symmetric encryption to protect media data, so any receiver also has the keys needed to encrypt packets for the sender 🔑👯.
7.3. Key Management 🗝️
the specifics of key management are beyond the scope of this document 🙃. still, every client SHOULD change their keys when new clients join or leave the call, for forward secrecy and post-compromise security 🔄💖.
7.4. Replay 🔁
the handling of replay is out of scope for this document 🙅♀️. however, senders MUST reject requests to encrypt multiple times with the same key and nonce, since several AEAD algorithms fail badly in those cases (see, e.g., Section 5.1.1 of [RFC5116]) 💀. this one is a hard no, periodt 🚫.
7.5. Risks Due to Short Tags ⚠️
the SFrame cipher suites based on AES-CTR allow for short authentication tags 🏷️. the catch is that short tags bring a higher risk that an attacker can make an SFrame receiver accept an SFrame ciphertext of the attacker's own choosing 😰.
assuming the cipher suite's authentication properties are robust, the only attack an attacker can mount is trying to find an acceptable (ciphertext, tag) combination through brute force 🎲. the expected success rate of that brute-force attack looks like this:
attacker_success_rate = attempts_per_second / 2^(8*Nt)for example, a gigabit Ethernet connection can transmit roughly 2^20 packets per second 🚀. if an attacker saturated that link with guesses against a 32-bit authentication tag (Nt=4), they'd succeed on average roughly once every 2^12 seconds, or about once an hour ⏰. yikes 😭.
in a typical SFrame usage in a real-time media application, there are a few ways to mitigate this risk 🛠️:
- Receivers only accept SFrame ciphertexts over HBH-secure channels (e.g., SRTP security associations or QUIC connections). If that's the case, only an entity that is part of such a channel can mount the above attack 🔐.
- The expected packet rate for a media stream is very predictable (and typically far lower than the example above). On one hand, attacks at this rate succeed even less often than the high-rate attack described above 📉. On the other hand, the application can treat an elevated packet arrival rate as a signal of a brute-force attack 🚨. this latter approach is common in other settings, like mitigating brute-force attacks on passwords 🔑.
- Media applications typically don't give feedback to media senders about which media packets failed to decrypt 🙊. when media-quality feedback mechanisms are used, decryption failures typically show up as packet losses, but only at an aggregate level 📊.
- Anti-replay mechanisms (see Section 7.4) stop the attacker from reusing valid ciphertexts (either observed or guessed) 🛑. a receiver applying anti-replay controls only accepts one valid plaintext per CTR value. since the CTR value is covered by SFrame authentication, the attacker has to do a fresh search for a valid tag for every forged ciphertext, even if the encrypted content stays the same 🔍. so when the brute-force attack succeeds, it only lets the attacker send a single SFrame ciphertext 🫠. that ciphertext can't be reused, because either it has the same CTR value and gets discarded as a replay, or it has a different CTR value and its tag is no longer valid 💅.
without these mitigations, an application that uses short tags is at heightened risk of forgery attacks 😱. in many cases it's simpler to use full-size tags and tolerate slightly higher bandwidth usage, rather than add the extra defenses needed to use short tags safely 💖.
8. IANA Considerations 📋
IANA has created a new registry called "SFrame Cipher Suites" (Section 8.1) under the "SFrame" group registry heading 🎀.
8.1. SFrame Cipher Suites 🔐
the "SFrame Cipher Suites" registry lists identifiers for SFrame cipher suites as defined in Section 4.5 🗂️. the cipher suite field is two bytes wide, so the valid cipher suites are in the range 0x0000 to 0xFFFF 🔢. except as noted below, assignments are made via the Specification Required policy [RFC8126] ✨.
the registration template looks like this 📝:
- Value: The numeric value of the cipher suite
- Name: The name of the cipher suite
- Recommended: Whether support for this cipher suite is recommended by the IETF. Valid values are "Y", "N", and "D" as described in Section 17.1 of [MLS-PROTO]. The default value of the "Recommended" column is "N". Setting the Recommended item to "Y" or "D", or changing an item whose current value is "Y" or "D", requires Standards Action [RFC8126] 🙌.
- Reference: The document where this cipher suite is defined
- Change Controller: Who is authorized to update the row in the registry
initial contents 🌸:
+========+============================+===+===========+============+
| Value | Name | R | Reference | Change |
| | | | | Controller |
+========+============================+===+===========+============+
| 0x0000 | Reserved | - | RFC 9605 | IETF |
+--------+----------------------------+---+-----------+------------+
| 0x0001 | AES_128_CTR_HMAC_SHA256_80 | Y | RFC 9605 | IETF |
+--------+----------------------------+---+-----------+------------+
| 0x0002 | AES_128_CTR_HMAC_SHA256_64 | Y | RFC 9605 | IETF |
+--------+----------------------------+---+-----------+------------+
| 0x0003 | AES_128_CTR_HMAC_SHA256_32 | Y | RFC 9605 | IETF |
+--------+----------------------------+---+-----------+------------+
| 0x0004 | AES_128_GCM_SHA256_128 | Y | RFC 9605 | IETF |
+--------+----------------------------+---+-----------+------------+
| 0x0005 | AES_256_GCM_SHA512_128 | Y | RFC 9605 | IETF |
+--------+----------------------------+---+-----------+------------+
| 0xF000 | Reserved for Private Use | - | RFC 9605 | IETF |
| - | | | | |
| 0xFFFF | | | | |
+--------+----------------------------+---+-----------+------------+Table 2: SFrame Cipher Suites 🎀
9. Application Responsibilities 🎀
Okay bestie, so to use SFrame, your app has to decide what goes into the encryption and decryption operations and how the SFrame ciphertexts get from sender to receiver, including any fragmenting and reassembling along the way 📦. This section is the checklist of extra stuff the app has to handle so SFrame stays secure 🔐💖
Generally, the app is the one in charge of setting up SFrame 👑. First it decides when SFrame gets applied at all. When it does get applied, the app picks the cipher suite. And if new SFrame versions show up later, the app is the one who figures out which version to use ✨
This split is pretty similar to how other media settings like codecs usually work in media apps. They get set up in some signaling protocol, not inside the media itself 🎬 Apps might find it handy to extend the protocols they already use to negotiate other media settings, like Session Description Protocol (SDP) [RFC8866], so they can negotiate SFrame settings too 🌸
9.1. Header Value Uniqueness 🔑
Applications MUST ensure that each (base_key, KID, CTR) combination is used for at most one SFrame encryption operation 🚫 This is what keeps the (key, nonce) pairs used by the underlying AEAD algorithm from ever getting reused, which is literally the whole point 😭
The usual move is to give each sender a KID or a set of KIDs. Then each sender uses the CTR field as a monotonic counter, bumping it up for every plaintext they encrypt 🔢 It's simple, and it keeps CTR values as small as possible, so there's less overhead. Iconic ✨
Now, if an SFrame context might get written to persistent storage, that context has to include the last-used CTR value 💾 Later on, when the context gets used again, the app should use the stored CTR value to figure out the next CTR value for an encryption operation. Then it writes that next CTR value back to storage before actually using it for encryption. Storing it before usage (instead of after) is what makes sure a storage failure can't cause the same (base_key, KID, CTR) combo to get reused 🛡️ Main character energy for sure 💅
9.2. Key Management Framework 🗝️
The app is responsible for provisioning SFrame with a mapping of KID values to base_key values, plus the resulting keys and salts 📋 Even more important, the app decides which KID values get used for which purposes, like which senders get which ones 🫶 An app's KID assignment strategy MUST be structured so that the non-reuse properties from Section 9.1 actually hold ✨
The app also owns the rotation schedule for keys 🔄 For example, one app might have an ephemeral group for every call and keep rotating keys whenever endpoints join or leave the call. Another app could have a persistent group that gets reused across multiple calls and just derives ephemeral symmetric keys for each specific call 💜 Both are valid vibes, just different ones 🌈
It should be noted that KID values are not encrypted by SFrame, which means any application-layer intermediaries that handle an SFrame ciphertext can see them 👀 Ngl this is a big deal: if there's any application meaning packed into KID values, that meaning gets exposed to those intermediaries. For example, with the scheme in Section 5.1, the number of ratchet steps per sender is exposed. And with the scheme in Section 5.2, the number of epochs and the MLS sender ID of the SFrame sender are exposed 🙈 Keep your KIDs boring, bestie 🫶
9.3. Anti-Replay 🔁
Handling anti-replay is the application's job 🙌 Replays by network attackers are assumed to be stopped by network-layer stuff, like TLS or SRTP 🛡️ And as mentioned in Section 7.4, senders MUST reject requests to encrypt multiple times with the same key and nonce 🚫 Absolutely no exceptions, periodt 💀
Anti-replay on the receiver side isn't mandatory to implement 🤷♀️ Receivers MAY apply time- or counter-based anti-replay mitigations if they want 🦋 For example, Section 3.3.2 of [RFC3711] specifies a counter-based anti-replay mitigation. That could be adapted for SFrame by using the CTR field as the counter 🧚
9.4. Metadata 🧾
The metadata input to SFrame operations is an opaque byte string that the application specifies 🫥 So the app has to decide what info goes into the metadata and make sure it gets handed to the encryption and decryption functions at the right points 📍 A receiver MUST NOT use SFrame-authenticated metadata until after the SFrame decrypt function has authenticated it. The one exception is when the goal is just to prepare an SFrame ciphertext for SFrame decryption 🙅♀️ Basically, metadata can be used "upstream of SFrame" in a processing pipeline, but only to get ready for SFrame decryption 🌸
Here's an example. Say an app uses SFrame to encrypt audio frames that get sent over SRTP, and some app data rides along in the RTP header extension 🎧 Now say the app also puts that same app data in the SFrame metadata, so the SFU can read it but not change it 👀 A receiver can use the app data in the RTP header extension as part of the normal SRTP decryption process, since that's required to get the SFrame ciphertext out of the SRTP payload 📦 But the receiver MUST NOT use the app data for any other purpose before SFrame decryption has authenticated it 🚨 Do not skip the line, bestie 😤
10. References 📚
10.1. Normative References 💅
[MLS-PROTO]
Barnes, R., Beurdouche, B., Robert, R., Millican, J.,
Omara, E., and K. Cohn-Gordon, "The Messaging Layer
Security (MLS) Protocol", RFC 9420, DOI 10.17487/RFC9420,
July 2023, <https://www.rfc-editor.org/info/rfc9420>.
[RFC2119] Bradner, S., "Key words for use in RFCs to Indicate
Requirement Levels", BCP 14, RFC 2119,
DOI 10.17487/RFC2119, March 1997,
<https://www.rfc-editor.org/info/rfc2119>.
[RFC5116] McGrew, D., "An Interface and Algorithms for Authenticated
Encryption", RFC 5116, DOI 10.17487/RFC5116, January 2008,
<https://www.rfc-editor.org/info/rfc5116>.
[RFC5869] Krawczyk, H. and P. Eronen, "HMAC-based Extract-and-Expand
Key Derivation Function (HKDF)", RFC 5869,
DOI 10.17487/RFC5869, May 2010,
<https://www.rfc-editor.org/info/rfc5869>.
[RFC8126] Cotton, M., Leiba, B., and T. Narten, "Guidelines for
Writing an IANA Considerations Section in RFCs", BCP 26,
RFC 8126, DOI 10.17487/RFC8126, June 2017,
<https://www.rfc-editor.org/info/rfc8126>.
[RFC8174] Leiba, B., "Ambiguity of Uppercase vs Lowercase in RFC
2119 Key Words", BCP 14, RFC 8174, DOI 10.17487/RFC8174,
May 2017, <https://www.rfc-editor.org/info/rfc8174>.10.2. Informative References 🦋
[MLS-ARCH] Beurdouche, B., Rescorla, E., Omara, E., Inguva, S., and
A. Duric, "The Messaging Layer Security (MLS)
Architecture", Work in Progress, Internet-Draft, draft-
ietf-mls-architecture-15, 3 August 2024,
<https://datatracker.ietf.org/doc/html/draft-ietf-mls-
architecture-15>.
[MOQ-TRANSPORT]
Curley, L., Pugin, K., Nandakumar, S., Vasiliev, V., and
I. Swett, Ed., "Media over QUIC Transport", Work in
Progress, Internet-Draft, draft-ietf-moq-transport-05, 8
July 2024, <https://datatracker.ietf.org/doc/html/draft-
ietf-moq-transport-05>.
[RFC3711] Baugher, M., McGrew, D., Naslund, M., Carrara, E., and K.
Norrman, "The Secure Real-time Transport Protocol (SRTP)",
RFC 3711, DOI 10.17487/RFC3711, March 2004,
<https://www.rfc-editor.org/info/rfc3711>.
[RFC6716] Valin, JM., Vos, K., and T. Terriberry, "Definition of the
Opus Audio Codec", RFC 6716, DOI 10.17487/RFC6716,
September 2012, <https://www.rfc-editor.org/info/rfc6716>.
[RFC7656] Lennox, J., Gross, K., Nandakumar, S., Salgueiro, G., and
B. Burman, Ed., "A Taxonomy of Semantics and Mechanisms
for Real-Time Transport Protocol (RTP) Sources", RFC 7656,
DOI 10.17487/RFC7656, November 2015,
<https://www.rfc-editor.org/info/rfc7656>.
[RFC7667] Westerlund, M. and S. Wenger, "RTP Topologies", RFC 7667,
DOI 10.17487/RFC7667, November 2015,
<https://www.rfc-editor.org/info/rfc7667>.
[RFC8723] Jennings, C., Jones, P., Barnes, R., and A.B. Roach,
"Double Encryption Procedures for the Secure Real-Time
Transport Protocol (SRTP)", RFC 8723,
DOI 10.17487/RFC8723, April 2020,
<https://www.rfc-editor.org/info/rfc8723>.
[RFC8866] Begen, A., Kyzivat, P., Perkins, C., and M. Handley, "SDP:
Session Description Protocol", RFC 8866,
DOI 10.17487/RFC8866, January 2021,
<https://www.rfc-editor.org/info/rfc8866>.
[RTP-PAYLOAD]
Murillo, S. G., Fablet, Y., and A. Gouaillard, "Codec
agnostic RTP payload format for video", Work in Progress,
Internet-Draft, draft-gouaillard-avtcore-codec-agn-rtp-
payload-01, 9 March 2021,
<https://datatracker.ietf.org/doc/html/draft-gouaillard-
avtcore-codec-agn-rtp-payload-01>.
[TestVectors]
"SFrame Test Vectors", commit 025d568, September 2023,
<https://github.com/sframe-wg/sframe/blob/025d568/test-
vectors/test-vectors.json>.
[WEBTRANSPORT]
Vasiliev, V., "The WebTransport Protocol Framework", Work
in Progress, Internet-Draft, draft-ietf-webtrans-overview-
08, 25 August 2024,
<https://datatracker.ietf.org/api/v1/doc/document/draft-
ietf-webtrans-overview/>.Appendix A. Example API ✨
This section is not normative. 🫶
okay so this section is a notional API, like a sketch of what an SFrame implementation might expose 💅 the big idea is the "SFrame context," and inside it KID values actually mean something. in the key management scheme from Section 5.1, every sender gets their own context 🧚 but in the scheme from Section 5.2, all the senders share one context. bestie, that difference is everything 😭
an SFrame context stores mappings from KID values to "key contexts," and the key context depends on whether the KID is for sending or receiving 🎀 (an SFrame key should never be used for both operations, so don't mix them up!!). a key context tracks the key and salt tied to the KID, plus the current CTR value. a sending key context also tracks the next CTR value it's going to use 🦋
here are the main moves you can make on an SFrame context:
- Create an SFrame context: The context is initialized with a cipher suite and no KID mappings. 🌸
- Add a key for sending: The key and salt are derived from the base key and used to initialize a send context, together with a zero CTR value. 💖
- Add a key for receiving: The key and salt are derived from the base key and used to initialize a send context. 👑
- Encrypt a plaintext: Encrypt a given plaintext using the key for a given KID, including the specified metadata. ✨
- Decrypt an SFrame ciphertext: Decrypt an SFrame ciphertext with the KID and CTR values specified in the SFrame header, and the provided metadata. 🔥
Figure 10 shows an example of the kinds of structures and methods you could use to build an SFrame API in Rust. it's giving clean and simple 🌈
type KeyId = u64;
type Counter = u64;
type CipherSuite = u16;
struct SendKeyContext {
key: Vec<u8>,
salt: Vec<u8>,
next_counter: Counter,
}
struct RecvKeyContext {
key: Vec<u8>,
salt: Vec<u8>,
}
struct SFrameContext {
cipher_suite: CipherSuite,
send_keys: HashMap<KeyId, SendKeyContext>,
recv_keys: HashMap<KeyId, RecvKeyContext>,
}
trait SFrameContextMethods {
fn create(cipher_suite: CipherSuite) -> Self;
fn add_send_key(&self, kid: KeyId, base_key: &[u8]);
fn add_recv_key(&self, kid: KeyId, base_key: &[u8]);
fn encrypt(&mut self, kid: KeyId, metadata: &[u8],
plaintext: &[u8]) -> Vec<u8>;
fn decrypt(&self, metadata: &[u8], ciphertext: &[u8]) -> Vec<u8>;
}
Figure 10: An Example SFrame APIAppendix B. Overhead Analysis
okay now the spicy math 🙈 any use of SFrame adds overhead, meaning extra bandwidth to send a given media stream. how much depends on a few things:
- The number of senders involved in a conference (length of KID) 🫶
- The duration of the conference (length of CTR) ⏳
- The cipher suite in use (length of authentication tag) 🔐
- Whether SFrame is used to encrypt packets, whole frames, or some other unit 📦
overall, you can estimate the overhead rate in kilobits per second like this:
OverheadKbps = (1 + |CTR| + |KID| + |TAG|) * 8 * CTPerSecond / 1024the constant 1 is the fixed SFrame header 🧁. |CTR| and |KID| are the lengths of those fields, |TAG| is the cipher overhead, and CTPerSecond is how many SFrame ciphertexts get sent per second (like packets or frames per second). simple formula, big vibes ✨
we're going to run overhead estimates for a bunch of common scenarios in the rest of this section 💅
B.1. Assumptions
okay so in these calculations we make conservative assumptions about SFrame overhead 🛡️ the goal is that the numbers we compute here are likely to be an upper bound on what you'd see in practice. main character planning 👑
+==============+=======+============================+
| Field | Bytes | Explanation |
+==============+=======+============================+
| Config byte | 1 | Fixed |
+--------------+-------+----------------------------+
| Key ID (KID) | 2 | >255 senders; or MLS epoch |
| | | (E=4) and >16 senders |
+--------------+-------+----------------------------+
| Counter | 3 | More than 24 hours of |
| (CTR) | | media in common cases |
+--------------+-------+----------------------------+
| Cipher | 16 | Full authentication tag |
| overhead | | (longest defined here) |
+--------------+-------+----------------------------+
Table 3: Overhead Analysis Assumptionsso in total, each SFrame encryption adds 22 bytes of overhead 🌸 we're gonna look at two scenarios: applying SFrame per frame and applying it per packet. in each one, we compute the SFrame overhead in absolute terms (kbps) and as a percentage of the base bandwidth 📊
B.2. Audio
in audio streams there's typically a one-to-one relationship between frames and packets 🎵 so the overhead is the same whether you use SFrame per packet or per frame. no difference, no drama 💜
Table 4 covers three scenarios based on the recommended configurations of the Opus codec [RFC6716] (where "fps" stands for "frames per second") 🎧
+==============+==============+=====+======+==========+==========+
| Scenario | Frame length | fps | Base | Overhead | Overhead |
| | | | kbps | kbps | % |
+==============+==============+=====+======+==========+==========+
| Narrow-band | 120 ms | 8.3 | 8 | 1.4 | 17.9% |
| speech | | | | | |
+--------------+--------------+-----+------+----------+----------+
| Full-band | 20 ms | 50 | 32 | 8.6 | 26.9% |
| speech | | | | | |
+--------------+--------------+-----+------+----------+----------+
| Full-band | 10 ms | 100 | 128 | 17.2 | 13.4% |
| stereo music | | | | | |
+--------------+--------------+-----+------+----------+----------+
Table 4: SFrame Overhead for Audio StreamsB.3. Video
okay video frames can be bigger than an MTU, so they're commonly split across multiple packets 📺 Tables 5 and 6 show the estimated overhead of encrypting a video stream. Table 5 is for SFrame applied per frame, and Table 6 is for SFrame applied per packet. the resolutions, frames per second, and bandwidths were picked to roughly reflect what modern video codecs can do, from very low to very high quality 🌈
+=============+=====+===========+===============+============+
| Scenario | fps | Base kbps | Overhead kbps | Overhead % |
+=============+=====+===========+===============+============+
| 426 x 240 | 7.5 | 45 | 1.3 | 2.9% |
+-------------+-----+-----------+---------------+------------+
| 640 x 360 | 15 | 200 | 2.6 | 1.3% |
+-------------+-----+-----------+---------------+------------+
| 640 x 360 | 30 | 400 | 5.2 | 1.3% |
+-------------+-----+-----------+---------------+------------+
| 1280 x 720 | 30 | 1500 | 5.2 | 0.3% |
+-------------+-----+-----------+---------------+------------+
| 1920 x 1080 | 60 | 7200 | 10.3 | 0.1% |
+-------------+-----+-----------+---------------+------------+
Table 5: SFrame Overhead for a Video Stream Encrypted per
Frame +==========+=====+==============+======+==========+==========+
| Scenario | fps | Packets per | Base | Overhead | Overhead |
| | | Second (pps) | kbps | kbps | % |
+==========+=====+==============+======+==========+==========+
| 426 x | 7.5 | 7.5 | 45 | 1.3 | 2.9% |
| 240 | | | | | |
+----------+-----+--------------+------+----------+----------+
| 640 x | 15 | 30 | 200 | 5.2 | 2.6% |
| 360 | | | | | |
+----------+-----+--------------+------+----------+----------+
| 640 x | 30 | 60 | 400 | 10.3 | 2.6% |
| 360 | | | | | |
+----------+-----+--------------+------+----------+----------+
| 1280 x | 30 | 180 | 1500 | 30.9 | 2.1% |
| 720 | | | | | |
+----------+-----+--------------+------+----------+----------+
| 1920 x | 60 | 780 | 7200 | 134.1 | 1.9% |
| 1080 | | | | | |
+----------+-----+--------------+------+----------+----------+
Table 6: SFrame Overhead for a Video Stream Encrypted per
Packetthe per-frame case is giving "the more quality you have, the less it hurts" 🥹 the SFrame percentage overhead heads toward zero as video quality improves, since bandwidth is driven more by picture size than by frame rate. in the per-packet case, the SFrame percentage overhead approaches the ratio of SFrame overhead per packet to the MTU. here that's 22 bytes of SFrame overhead divided by an assumed 1200-byte MTU, which is about 1.8% 🧮
B.4. Conferences
real conferences usually have several audio and video streams running at once 🎉 the overhead of SFrame in a conference is the total of the overhead across all those individual streams. so yes, SFrame has a big percentage overhead on an audio stream, but if the conference also has a video stream, the audio overhead is probably negligible next to the conference's overall bandwidth 🫧
for example, Table 7 shows the overhead estimates for a two-person conference where one person sends low-quality media and the other sends high-quality media. (we assume SFrame is applied per frame here.) the video streams dominate the bandwidth at the SFU, so the total bandwidth overhead is only around 1% 💀✨
+=====================+===========+===============+============+
| Stream | Base Kbps | Overhead Kbps | Overhead % |
+=====================+===========+===============+============+
| Participant 1 audio | 8 | 1.4 | 17.9% |
+---------------------+-----------+---------------+------------+
| Participant 1 video | 45 | 1.3 | 2.9% |
+---------------------+-----------+---------------+------------+
| Participant 2 audio | 32 | 9 | 26.9% |
+---------------------+-----------+---------------+------------+
| Participant 2 video | 1500 | 5 | 0.3% |
+---------------------+-----------+---------------+------------+
| Total at SFU | 1585 | 16.5 | 1.0% |
+---------------------+-----------+---------------+------------+
Table 7: SFrame Overhead for a Two-Person ConferenceB.5. SFrame over RTP
okay so bestie, this part is where SFrame meets RTP 🦋 and it's giving "how do we make these two actually vibe together" energy 💅
SFrame is a generic encapsulation format, but a lot of the apps that are likely to use it are built on RTP 🎀 This section talks about how you could integrate SFrame and RTP, and what challenges would need to be overcome 💖
As we covered in Section 4.1, there are two natural patterns for plugging SFrame into an app: applying SFrame per frame or per packet ✨ For RTP-based apps, per packet means that the payload of each RTP packet is an SFrame ciphertext that starts with an SFrame header, as shown in Figure 11 🌸 Per frame means different RTP payloads look different: the first payload of a frame carries the SFrame headers, and the later payloads carry more chunks of the ciphertext, as shown in Figure 12 🧚
Receivers need to know which of these schemes the sender used for a given sequence of RTP packets, or they won't interpret the media payloads properly 😭 SFrame does NOT give you a way to distribute this configuration info, though 💀 For apps that use SDP to negotiate RTP media streams [RFC8866], an appropriate SDP extension could handle this job 🫶
Per-frame application also needs packetization and depacketization to be done in a generic way that doesn't depend on the media content of the packets 🙈 That's because what's being packetized is opaque ciphertext (except for the SFrame header) 🔐 For this generic packetization to work interoperably, one would have to be defined, for example as proposed in [RTP-PAYLOAD] 👑
Here's the RTP packet layout with an SFrame-protected payload 🦋
+---+-+-+-------+-+-----------+------------------------------+<-+
|V=2|P|X| CC |M| PT | sequence number | |
+---+-+-+-------+-+-----------+------------------------------+ |
| timestamp | |
+------------------------------------------------------------+ |
| synchronization source (SSRC) identifier | |
+============================================================+ |
| contributing source (CSRC) identifiers | |
| .... | |
+------------------------------------------------------------+ |
| RTP extension(s) (OPTIONAL) | |
+->+-------------------+----------------------------------------+ |
| | SFrame header | | |
| +-------------------+ | |
| | | |
| | SFrame encrypted and authenticated payload | |
| | | |
+->+------------------------------------------------------------+<-+
| | SRTP authentication tag | |
| +------------------------------------------------------------+ |
| |
+--- SRTP Encrypted Portion SRTP Authenticated Portion ---+
Figure 11: SRTP Packet with SFrame-Protected PayloadAnd here's the per-frame encryption flow for RTP 🌈
+----------------+ +---------------+
| frame metadata | | |
+-------+--------+ | |
| | frame |
| | |
| | |
| +-------+-------+
| |
| |
V V
+--------------------------------------+
| SFrame Encrypt |
+--------------------------------------+
| |
| |
| V
| +-------+-------+
| | |
| | |
| | encrypted |
| | frame |
| | |
| | |
| +-------+-------+
| |
| generic RTP packetize
| |
| +----------------------+--------.....--------+
| | | |
V V V V
+---------------+ +---------------+ +---------------+
| SFrame header | | | | |
+---------------+ | | | |
| | | payload 2/N | ... | payload N/N |
| payload 1/N | | | | |
| | | | | |
+---------------+ +---------------+ +---------------+
Figure 12: Encryption Flow with per-Frame Encryption for RTPAppendix C. Test Vectors 🧪
okay this is the part where implementers get to check their homework, and I'm obsessed 😭 This section gives you test vectors so you can verify that your SFrame encryption and decryption is actually correct 💅 Besides vectors for the whole encrypt/decrypt process, there are also vectors for header encoding/decoding, and for AEAD encryption/decryption using the AES-CTR construction defined in Section 4.5.1 ✨
All values are either numeric or byte strings 🍓 Numbers are written as hex with a 0x prefix, and byte strings are written in hex encoding 🌸
Line breaks and whitespace inside values are added so the text fits the RFC width requirements, so take them out before you use the values 🙈 The same vectors are also available in JSON at [TestVectors] 💖 In the JSON version, numbers are JSON numbers and byte strings are JSON strings holding the hex encoding 🫶
C.1. Header Encoding/Decoding
For each case, there are three things:
- kid: A KID value
- ctr: A CTR value
- header: An encoded SFrame header
Your implementation should check that:
- Encoding a header with the KID and CTR gives you the provided header value
- Decoding the provided header value gives you the provided KID and CTR values
Here are the cases 🎀
kid: 0x0000000000000000
ctr: 0x0000000000000000
header: 00
kid: 0x0000000000000000
ctr: 0x0000000000000001
header: 01
kid: 0x0000000000000000
ctr: 0x00000000000000ff
header: 08ff
kid: 0x0000000000000000
ctr: 0x0000000000000100
header: 090100
kid: 0x0000000000000000
ctr: 0x000000000000ffff
header: 09ffff
kid: 0x0000000000000000
ctr: 0x0000000000010000
header: 0a010000
kid: 0x0000000000000000
ctr: 0x0000000000ffffff
header: 0affffff
kid: 0x0000000000000000
ctr: 0x0000000001000000
header: 0b01000000
kid: 0x0000000000000000
ctr: 0x00000000ffffffff
header: 0bffffffff
kid: 0x0000000000000000
ctr: 0x0000000100000000
header: 0c0100000000
kid: 0x0000000000000000
ctr: 0x000000ffffffffff
header: 0cffffffffff
kid: 0x0000000000000000
ctr: 0x0000010000000000
header: 0d010000000000
kid: 0x0000000000000000
ctr: 0x0000ffffffffffff
header: 0dffffffffffff
kid: 0x0000000000000000
ctr: 0x0001000000000000
header: 0e01000000000000
kid: 0x0000000000000000
ctr: 0x00ffffffffffffff
header: 0effffffffffffff
kid: 0x0000000000000000
ctr: 0x0100000000000000
header: 0f0100000000000000
kid: 0x0000000000000000
ctr: 0xffffffffffffffff
header: 0fffffffffffffffff
kid: 0x0000000000000001
ctr: 0x0000000000000000
header: 10
kid: 0x0000000000000001
ctr: 0x0000000000000001
header: 11
kid: 0x0000000000000001
ctr: 0x00000000000000ff
header: 18ff
kid: 0x0000000000000001
ctr: 0x0000000000000100
header: 190100
kid: 0x0000000000000001
ctr: 0x000000000000ffff
header: 19ffff
kid: 0x0000000000000001
ctr: 0x0000000000010000
header: 1a010000
kid: 0x0000000000000001
ctr: 0x0000000000ffffff
header: 1affffff
kid: 0x0000000000000001
ctr: 0x0000000001000000
header: 1b01000000
kid: 0x0000000000000001
ctr: 0x00000000ffffffff
header: 1bffffffff
kid: 0x0000000000000001
ctr: 0x0000000100000000
header: 1c0100000000
kid: 0x0000000000000001
ctr: 0x000000ffffffffff
header: 1cffffffffff
kid: 0x0000000000000001
ctr: 0x0000010000000000
header: 1d010000000000
kid: 0x0000000000000001
ctr: 0x0000ffffffffffff
header: 1dffffffffffff
kid: 0x0000000000000001
ctr: 0x0001000000000000
header: 1e01000000000000
kid: 0x0000000000000001
ctr: 0x00ffffffffffffff
header: 1effffffffffffff
kid: 0x0000000000000001
ctr: 0x0100000000000000
header: 1f0100000000000000
kid: 0x0000000000000001
ctr: 0xffffffffffffffff
header: 1fffffffffffffffff
kid: 0x00000000000000ff
ctr: 0x0000000000000000
header: 80ff
kid: 0x00000000000000ff
ctr: 0x0000000000000001
header: 81ff
kid: 0x00000000000000ff
ctr: 0x00000000000000ff
header: 88ffff
kid: 0x00000000000000ff
ctr: 0x0000000000000100
header: 89ff0100
kid: 0x00000000000000ff
ctr: 0x000000000000ffff
header: 89ffffff
kid: 0x00000000000000ff
ctr: 0x0000000000010000
header: 8aff010000
kid: 0x00000000000000ff
ctr: 0x0000000000ffffff
header: 8affffffff
kid: 0x00000000000000ff
ctr: 0x0000000001000000
header: 8bff01000000
kid: 0x00000000000000ff
ctr: 0x00000000ffffffff
header: 8bffffffffff
kid: 0x00000000000000ff
ctr: 0x0000000100000000
header: 8cff0100000000
kid: 0x00000000000000ff
ctr: 0x000000ffffffffff
header: 8cffffffffffff
kid: 0x00000000000000ff
ctr: 0x0000010000000000
header: 8dff010000000000
kid: 0x00000000000000ff
ctr: 0x0000ffffffffffff
header: 8dffffffffffffff
kid: 0x00000000000000ff
ctr: 0x0001000000000000
header: 8eff01000000000000
kid: 0x00000000000000ff
ctr: 0x00ffffffffffffff
header: 8effffffffffffffff
kid: 0x00000000000000ff
ctr: 0x0100000000000000
header: 8fff0100000000000000
kid: 0x00000000000000ff
ctr: 0xffffffffffffffff
header: 8fffffffffffffffffff
kid: 0x0000000000000100
ctr: 0x0000000000000000
header: 900100
kid: 0x0000000000000100
ctr: 0x0000000000000001
header: 910100
kid: 0x0000000000000100
ctr: 0x00000000000000ff
header: 980100ff
kid: 0x0000000000000100
ctr: 0x0000000000000100
header: 9901000100
kid: 0x0000000000000100
ctr: 0x000000000000ffff
header: 990100ffff
kid: 0x0000000000000100
ctr: 0x0000000000010000
header: 9a0100010000
kid: 0x0000000000000100
ctr: 0x0000000000ffffff
header: 9a0100ffffff
kid: 0x0000000000000100
ctr: 0x0000000001000000
header: 9b010001000000
kid: 0x0000000000000100
ctr: 0x00000000ffffffff
header: 9b0100ffffffff
kid: 0x0000000000000100
ctr: 0x0000000100000000
header: 9c01000100000000
kid: 0x0000000000000100
ctr: 0x000000ffffffffff
header: 9c0100ffffffffff
kid: 0x0000000000000100
ctr: 0x0000010000000000
header: 9d0100010000000000
kid: 0x0000000000000100
ctr: 0x0000ffffffffffff
header: 9d0100ffffffffffff
kid: 0x0000000000000100
ctr: 0x0001000000000000
header: 9e010001000000000000
kid: 0x0000000000000100
ctr: 0x00ffffffffffffff
header: 9e0100ffffffffffffff
kid: 0x0000000000000100
ctr: 0x0100000000000000
header: 9f01000100000000000000
kid: 0x0000000000000100
ctr: 0xffffffffffffffff
header: 9f0100ffffffffffffffffOkay bestie, here are the rest of the header test vectors, straight from the source 🎀✨
kid: 0x0000000001000000
ctr: 0x0000000000000000
header: b001000000
kid: 0x0000000001000000
ctr: 0x0000000000000001
header: b101000000
kid: 0x0000000001000000
ctr: 0x00000000000000ff
header: b801000000ff
kid: 0x0000000001000000
ctr: 0x0000000000000100
header: b9010000000100
kid: 0x0000000001000000
ctr: 0x000000000000ffff
header: b901000000ffff
kid: 0x0000000001000000
ctr: 0x0000000000010000
header: ba01000000010000
kid: 0x0000000001000000
ctr: 0x0000000000ffffff
header: ba01000000ffffff
kid: 0x0000000001000000
ctr: 0x0000000001000000
header: bb0100000001000000
kid: 0x0000000001000000
ctr: 0x00000000ffffffff
header: bb01000000ffffffff
kid: 0x0000000001000000
ctr: 0x0000000100000000
header: bc010000000100000000
kid: 0x0000000001000000
ctr: 0x000000ffffffffff
header: bc01000000ffffffffff
kid: 0x0000000001000000
ctr: 0x0000010000000000
header: bd01000000010000000000
kid: 0x0000000001000000
ctr: 0x0000ffffffffffff
header: bd01000000ffffffffffff
kid: 0x0000000001000000
ctr: 0x0001000000000000
header: be0100000001000000000000
kid: 0x0000000001000000
ctr: 0x00ffffffffffffff
header: be01000000ffffffffffffff
kid: 0x0000000001000000
ctr: 0x0100000000000000
header: bf010000000100000000000000
kid: 0x0000000001000000
ctr: 0xffffffffffffffff
header: bf01000000ffffffffffffffff
kid: 0x00000000ffffffff
ctr: 0x0000000000000000
header: b0ffffffff
kid: 0x00000000ffffffff
ctr: 0x0000000000000001
header: b1ffffffff
kid: 0x00000000ffffffff
ctr: 0x00000000000000ff
header: b8ffffffffff
kid: 0x00000000ffffffff
ctr: 0x0000000000000100
header: b9ffffffff0100
kid: 0x00000000ffffffff
ctr: 0x000000000000ffff
header: b9ffffffffffff
kid: 0x00000000ffffffff
ctr: 0x0000000000010000
header: baffffffff010000
kid: 0x00000000ffffffff
ctr: 0x0000000000ffffff
header: baffffffffffffff
kid: 0x00000000ffffffff
ctr: 0x0000000001000000
header: bbffffffff01000000
kid: 0x00000000ffffffff
ctr: 0x00000000ffffffff
header: bbffffffffffffffff
kid: 0x00000000ffffffff
ctr: 0x0000000100000000
header: bcffffffff0100000000
kid: 0x00000000ffffffff
ctr: 0x000000ffffffffff
header: bcffffffffffffffffff
kid: 0x00000000ffffffff
ctr: 0x0000010000000000
header: bdffffffff010000000000
kid: 0x00000000ffffffff
ctr: 0x0000ffffffffffff
header: bdffffffffffffffffffff
kid: 0x00000000ffffffff
ctr: 0x0001000000000000
header: beffffffff01000000000000
kid: 0x00000000ffffffff
ctr: 0x00ffffffffffffff
header: beffffffffffffffffffffff
kid: 0x00000000ffffffff
ctr: 0x0100000000000000
header: bfffffffff0100000000000000
kid: 0x00000000ffffffff
ctr: 0xffffffffffffffff
header: bfffffffffffffffffffffffff
kid: 0x0000000100000000
ctr: 0x0000000000000000
header: c00100000000
kid: 0x0000000100000000
ctr: 0x0000000000000001
header: c10100000000
kid: 0x0000000100000000
ctr: 0x00000000000000ff
header: c80100000000ff
kid: 0x0000000100000000
ctr: 0x0000000000000100
header: c901000000000100
kid: 0x0000000100000000
ctr: 0x000000000000ffff
header: c90100000000ffff
kid: 0x0000000100000000
ctr: 0x0000000000010000
header: ca0100000000010000
kid: 0x0000000100000000
ctr: 0x0000000000ffffff
header: ca0100000000ffffff
kid: 0x0000000100000000
ctr: 0x0000000001000000
header: cb010000000001000000
kid: 0x0000000100000000
ctr: 0x00000000ffffffff
header: cb0100000000ffffffff
kid: 0x0000000100000000
ctr: 0x0000000100000000
header: cc01000000000100000000
kid: 0x0000000100000000
ctr: 0x000000ffffffffff
header: cc0100000000ffffffffff
kid: 0x0000000100000000
ctr: 0x0000010000000000
header: cd0100000000010000000000
kid: 0x0000000100000000
ctr: 0x0000ffffffffffff
header: cd0100000000ffffffffffff
kid: 0x0000000100000000
ctr: 0x0001000000000000
header: ce010000000001000000000000
kid: 0x0000000100000000
ctr: 0x00ffffffffffffff
header: ce0100000000ffffffffffffff
kid: 0x0000000100000000
ctr: 0x0100000000000000
header: cf01000000000100000000000000
kid: 0x0000000100000000
ctr: 0xffffffffffffffff
header: cf0100000000ffffffffffffffff
kid: 0x000000ffffffffff
ctr: 0x0000000000000000
header: c0ffffffffff
kid: 0x000000ffffffffff
ctr: 0x0000000000000001
header: c1ffffffffff
kid: 0x000000ffffffffff
ctr: 0x00000000000000ff
header: c8ffffffffffff
kid: 0x000000ffffffffff
ctr: 0x0000000000000100
header: c9ffffffffff0100
kid: 0x000000ffffffffff
ctr: 0x000000000000ffff
header: c9ffffffffffffff
kid: 0x000000ffffffffff
ctr: 0x0000000000010000
header: caffffffffff010000
kid: 0x000000ffffffffff
ctr: 0x0000000000ffffff
header: caffffffffffffffff
kid: 0x000000ffffffffff
ctr: 0x0000000001000000
header: cbffffffffff01000000
kid: 0x000000ffffffffff
ctr: 0x00000000ffffffff
header: cbffffffffffffffffff
kid: 0x000000ffffffffff
ctr: 0x0000000100000000
header: ccffffffffff0100000000
kid: 0x000000ffffffffff
ctr: 0x000000ffffffffff
header: ccffffffffffffffffffff
kid: 0x000000ffffffffff
ctr: 0x0000010000000000
header: cdffffffffff010000000000
kid: 0x000000ffffffffff
ctr: 0x0000ffffffffffff
header: cdffffffffffffffffffffff
kid: 0x000000ffffffffff
ctr: 0x0001000000000000
header: ceffffffffff01000000000000
kid: 0x000000ffffffffff
ctr: 0x00ffffffffffffff
header: ceffffffffffffffffffffffff
kid: 0x000000ffffffffff
ctr: 0x0100000000000000
header: cfffffffffff0100000000000000
kid: 0x000000ffffffffff
ctr: 0xffffffffffffffff
header: cfffffffffffffffffffffffffff
kid: 0x0000010000000000
ctr: 0x0000000000000000
header: d0010000000000
kid: 0x0000010000000000
ctr: 0x0000000000000001
header: d1010000000000
kid: 0x0000010000000000
ctr: 0x00000000000000ff
header: d8010000000000ff
kid: 0x0000010000000000
ctr: 0x0000000000000100
header: d90100000000000100
kid: 0x0000010000000000
ctr: 0x000000000000ffff
header: d9010000000000ffff
kid: 0x0000010000000000
ctr: 0x0000000000010000
header: da010000000000010000
kid: 0x0000010000000000
ctr: 0x0000000000ffffff
header: da010000000000ffffff
kid: 0x0000010000000000
ctr: 0x0000000001000000
header: db01000000000001000000
kid: 0x0000010000000000
ctr: 0x00000000ffffffff
header: db010000000000ffffffff
kid: 0x0000010000000000
ctr: 0x0000000100000000
header: dc0100000000000100000000
kid: 0x0000010000000000
ctr: 0x000000ffffffffff
header: dc010000000000ffffffffff
kid: 0x0000010000000000
ctr: 0x0000010000000000
header: dd010000000000010000000000
kid: 0x0000010000000000
ctr: 0x0000ffffffffffff
header: dd010000000000ffffffffffff
kid: 0x0000010000000000
ctr: 0x0001000000000000
header: de01000000000001000000000000
kid: 0x0000010000000000
ctr: 0x00ffffffffffffff
header: de010000000000ffffffffffffff
kid: 0x0000010000000000
ctr: 0x0100000000000000
header: df0100000000000100000000000000
kid: 0x0000010000000000
ctr: 0xffffffffffffffff
header: df010000000000ffffffffffffffff
kid: 0x0000ffffffffffff
ctr: 0x0000000000000000
header: d0ffffffffffff
kid: 0x0000ffffffffffff
ctr: 0x0000000000000001
header: d1ffffffffffff
kid: 0x0000ffffffffffff
ctr: 0x00000000000000ff
header: d8ffffffffffffff
kid: 0x0000ffffffffffff
ctr: 0x0000000000000100
header: d9ffffffffffff0100 kid: 0x0000ffffffffffff
ctr: 0x000000000000ffff
header: d9ffffffffffffffff
kid: 0x0000ffffffffffff
ctr: 0x0000000000010000
header: daffffffffffff010000
kid: 0x0000ffffffffffff
ctr: 0x0000000000ffffff
header: daffffffffffffffffff
kid: 0x0000ffffffffffff
ctr: 0x0000000001000000
header: dbffffffffffff01000000
kid: 0x0000ffffffffffff
ctr: 0x00000000ffffffff
header: dbffffffffffffffffffff
kid: 0x0000ffffffffffff
ctr: 0x0000000100000000
header: dcffffffffffff0100000000
kid: 0x0000ffffffffffff
ctr: 0x000000ffffffffff
header: dcffffffffffffffffffffff
kid: 0x0000ffffffffffff
ctr: 0x0000010000000000
header: ddffffffffffff010000000000
kid: 0x0000ffffffffffff
ctr: 0x0000ffffffffffff
header: ddffffffffffffffffffffffff
kid: 0x0000ffffffffffff
ctr: 0x0001000000000000
header: deffffffffffff01000000000000
kid: 0x0000ffffffffffff
ctr: 0x00ffffffffffffff
header: deffffffffffffffffffffffffff
kid: 0x0000ffffffffffff
ctr: 0x0100000000000000
header: dfffffffffffff0100000000000000
kid: 0x0000ffffffffffff
ctr: 0xffffffffffffffff
header: dfffffffffffffffffffffffffffff
kid: 0x0001000000000000
ctr: 0x0000000000000000
header: e001000000000000
kid: 0x0001000000000000
ctr: 0x0000000000000001
header: e101000000000000
kid: 0x0001000000000000
ctr: 0x00000000000000ff
header: e801000000000000ff
kid: 0x0001000000000000
ctr: 0x0000000000000100
header: e9010000000000000100
kid: 0x0001000000000000
ctr: 0x000000000000ffff
header: e901000000000000ffff
kid: 0x0001000000000000
ctr: 0x0000000000010000
header: ea01000000000000010000
kid: 0x0001000000000000
ctr: 0x0000000000ffffff
header: ea01000000000000ffffff
kid: 0x0001000000000000
ctr: 0x0000000001000000
header: eb0100000000000001000000
kid: 0x0001000000000000
ctr: 0x00000000ffffffff
header: eb01000000000000ffffffff
kid: 0x0001000000000000
ctr: 0x0000000100000000
header: ec010000000000000100000000
kid: 0x0001000000000000
ctr: 0x000000ffffffffff
header: ec01000000000000ffffffffff
kid: 0x0001000000000000
ctr: 0x0000010000000000
header: ed01000000000000010000000000
kid: 0x0001000000000000
ctr: 0x0000ffffffffffff
header: ed01000000000000ffffffffffff
kid: 0x0001000000000000
ctr: 0x0001000000000000
header: ee0100000000000001000000000000
kid: 0x0001000000000000
ctr: 0x00ffffffffffffff
header: ee01000000000000ffffffffffffff
kid: 0x0001000000000000
ctr: 0x0100000000000000
header: ef010000000000000100000000000000
kid: 0x0001000000000000
ctr: 0xffffffffffffffff
header: ef01000000000000ffffffffffffffff
kid: 0x00ffffffffffffff
ctr: 0x0000000000000000
header: e0ffffffffffffff
kid: 0x00ffffffffffffff
ctr: 0x0000000000000001
header: e1ffffffffffffff
kid: 0x00ffffffffffffff
ctr: 0x00000000000000ff
header: e8ffffffffffffffff
kid: 0x00ffffffffffffff
ctr: 0x0000000000000100
header: e9ffffffffffffff0100
kid: 0x00ffffffffffffff
ctr: 0x000000000000ffff
header: e9ffffffffffffffffff
kid: 0x00ffffffffffffff
ctr: 0x0000000000010000
header: eaffffffffffffff010000
kid: 0x00ffffffffffffff
ctr: 0x0000000000ffffff
header: eaffffffffffffffffffff
kid: 0x00ffffffffffffff
ctr: 0x0000000001000000
header: ebffffffffffffff01000000
kid: 0x00ffffffffffffff
ctr: 0x00000000ffffffff
header: ebffffffffffffffffffffff
kid: 0x00ffffffffffffff
ctr: 0x0000000100000000
header: ecffffffffffffff0100000000
kid: 0x00ffffffffffffff
ctr: 0x000000ffffffffff
header: ecffffffffffffffffffffffff
kid: 0x00ffffffffffffff
ctr: 0x0000010000000000
header: edffffffffffffff010000000000
kid: 0x00ffffffffffffff
ctr: 0x0000ffffffffffff
header: edffffffffffffffffffffffffff
kid: 0x00ffffffffffffff
ctr: 0x0001000000000000
header: eeffffffffffffff01000000000000
kid: 0x00ffffffffffffff
ctr: 0x00ffffffffffffff
header: eeffffffffffffffffffffffffffff
kid: 0x00ffffffffffffff
ctr: 0x0100000000000000
header: efffffffffffffff0100000000000000
kid: 0x00ffffffffffffff
ctr: 0xffffffffffffffff
header: efffffffffffffffffffffffffffffff
kid: 0x0100000000000000
ctr: 0x0000000000000000
header: f00100000000000000
kid: 0x0100000000000000
ctr: 0x0000000000000001
header: f10100000000000000
kid: 0x0100000000000000
ctr: 0x00000000000000ff
header: f80100000000000000ff
kid: 0x0100000000000000
ctr: 0x0000000000000100
header: f901000000000000000100
kid: 0x0100000000000000
ctr: 0x000000000000ffff
header: f90100000000000000ffff
kid: 0x0100000000000000
ctr: 0x0000000000010000
header: fa0100000000000000010000
kid: 0x0100000000000000
ctr: 0x0000000000ffffff
header: fa0100000000000000ffffff
kid: 0x0100000000000000
ctr: 0x0000000001000000
header: fb010000000000000001000000
kid: 0x0100000000000000
ctr: 0x00000000ffffffff
header: fb0100000000000000ffffffff
kid: 0x0100000000000000
ctr: 0x0000000100000000
header: fc01000000000000000100000000
kid: 0x0100000000000000
ctr: 0x000000ffffffffff
header: fc0100000000000000ffffffffff
kid: 0x0100000000000000
ctr: 0x0000010000000000
header: fd0100000000000000010000000000
kid: 0x0100000000000000
ctr: 0x0000ffffffffffff
header: fd0100000000000000ffffffffffff
kid: 0x0100000000000000
ctr: 0x0001000000000000
header: fe010000000000000001000000000000
kid: 0x0100000000000000
ctr: 0x00ffffffffffffff
header: fe0100000000000000ffffffffffffff
kid: 0x0100000000000000
ctr: 0x0100000000000000
header: ff010000000000000001000000000000
00
kid: 0x0100000000000000
ctr: 0xffffffffffffffff
header: ff0100000000000000ffffffffffffff
ff
kid: 0xffffffffffffffff
ctr: 0x0000000000000000
header: f0ffffffffffffffff
kid: 0xffffffffffffffff
ctr: 0x0000000000000001
header: f1ffffffffffffffff
kid: 0xffffffffffffffff
ctr: 0x00000000000000ff
header: f8ffffffffffffffffff
kid: 0xffffffffffffffff
ctr: 0x0000000000000100
header: f9ffffffffffffffff0100
kid: 0xffffffffffffffff
ctr: 0x000000000000ffff
header: f9ffffffffffffffffffff
kid: 0xffffffffffffffff
ctr: 0x0000000000010000
header: faffffffffffffffff010000
kid: 0xffffffffffffffff
ctr: 0x0000000000ffffff
header: faffffffffffffffffffffff
kid: 0xffffffffffffffff
ctr: 0x0000000001000000
header: fbffffffffffffffff01000000
kid: 0xffffffffffffffff
ctr: 0x00000000ffffffff
header: fbffffffffffffffffffffffff
kid: 0xffffffffffffffff
ctr: 0x0000000100000000
header: fcffffffffffffffff0100000000
kid: 0xffffffffffffffff
ctr: 0x000000ffffffffff
header: fcffffffffffffffffffffffffff
kid: 0xffffffffffffffff
ctr: 0x0000010000000000
header: fdffffffffffffffff010000000000
kid: 0xffffffffffffffff
ctr: 0x0000ffffffffffff
header: fdffffffffffffffffffffffffffff
kid: 0xffffffffffffffff
ctr: 0x0001000000000000
header: feffffffffffffffff01000000000000
kid: 0xffffffffffffffff
ctr: 0x00ffffffffffffff
header: feffffffffffffffffffffffffffffff
kid: 0xffffffffffffffff
ctr: 0x0100000000000000
header: ffffffffffffffffff01000000000000
00
kid: 0xffffffffffffffff
ctr: 0xffffffffffffffff
header: ffffffffffffffffffffffffffffffff
ffC.2. AEAD Encryption/Decryption Using AES-CTR and HMAC ✨
okay bestie, now we get to the AEAD test vectors 🎀 so for each case we give you a bunch of little labelled values so you can check your work:
cipher_suite: the index of the cipher suite in use (see Section 8.1) 💅key: the key input to encryption/decryption 🔑enc_key: the encryption subkey produced by thederive_subkeys()algorithm 🌸auth_key: the encryption subkey produced by thederive_subkeys()algorithm 🦋nonce: the nonce input to encryption/decryption 🧚aad: the aad input to encryption/decryption 💜pt: the plaintext 📝ct: the ciphertext 🔥
so an implementation should verify that the following are true, where AEAD.Encrypt and AEAD.Decrypt are as defined in Section 4.5.1, and this is the whole vibe check:
AEAD.Encrypt(key, nonce, aad, pt) == ct✨AEAD.Decrypt(key, nonce, aad, ct) == pt✨
the other values in the test vector are just intermediate values that are there to help you debug when something goes sideways 🙈
cipher_suite: 0x0001
key: 000102030405060708090a0b0c0d0e0f
101112131415161718191a1b1c1d1e1f
202122232425262728292a2b2c2d2e2f
enc_key: 000102030405060708090a0b0c0d0e0f
auth_key: 101112131415161718191a1b1c1d1e1f
202122232425262728292a2b2c2d2e2f
nonce: 101112131415161718191a1b
aad: 4945544620534672616d65205747
pt: 64726166742d696574662d736672616d
652d656e63
ct: 6339af04ada1d064688a442b8dc69d5b
6bfa40f4bef0583e8081069cc60705
cipher_suite: 0x0002
key: 000102030405060708090a0b0c0d0e0f
101112131415161718191a1b1c1d1e1f
202122232425262728292a2b2c2d2e2f
enc_key: 000102030405060708090a0b0c0d0e0f
auth_key: 101112131415161718191a1b1c1d1e1f
202122232425262728292a2b2c2d2e2f
nonce: 101112131415161718191a1b
aad: 4945544620534672616d65205747
pt: 64726166742d696574662d736672616d
652d656e63
ct: 6339af04ada1d064688a442b8dc69d5b
6bfa40f4be6e93b7da076927bb
cipher_suite: 0x0003
key: 000102030405060708090a0b0c0d0e0f
101112131415161718191a1b1c1d1e1f
202122232425262728292a2b2c2d2e2f
enc_key: 000102030405060708090a0b0c0d0e0f
auth_key: 101112131415161718191a1b1c1d1e1f
202122232425262728292a2b2c2d2e2f
nonce: 101112131415161718191a1b
aad: 4945544620534672616d65205747
pt: 64726166742d696574662d736672616d
652d656e63
ct: 6339af04ada1d064688a442b8dc69d5b
6bfa40f4be09480509C.3. SFrame Encryption/Decryption ✨
okay so for each case, we give you a whole bundle of goodies 🎀:
cipher_suite: The index of the cipher suite in use (see Section 8.1)kid: A KID valuectr: A CTR valuebase_key: The base_key input to the derive_key_salt algorithmsframe_key_label: The label used to derive sframe_key in the derive_key_salt algorithmsframe_salt_label: The label used to derive sframe_salt in the derive_key_salt algorithmsframe_secret: The sframe_secret variable in the derive_key_salt algorithmsframe_key: The sframe_key value produced by the derive_key_salt algorithmsframe_salt: The sframe_salt value produced by the derive_key_salt algorithmmetadata: The metadata input to the SFrame encrypt algorithmpt: The plaintextct: The SFrame ciphertext
An implementation should verify that the following are true, where encrypt and decrypt are as defined in Section 4.4, using an SFrame context initialized with base_key assigned to kid 🫶:
encrypt(ctr, kid, metadata, plaintext) == ctdecrypt(metadata, ct) == pt
The other values in the test vector are intermediate values provided to facilitate debugging of test failures. Think of them as the breadcrumbs 🍓 that help you find where things went sideways 😭.
Vector one, with cipher_suite 0x0001 💜:
cipher_suite: 0x0001
kid: 0x0000000000000123
ctr: 0x0000000000004567
base_key: 000102030405060708090a0b0c0d0e0f
sframe_key_label: 534672616d6520312e30205365637265
74206b65792000000000000001230001
sframe_salt_label: 534672616d6520312e30205365637265
742073616c7420000000000000012300
01
sframe_secret: d926952ca8b7ec4a95941d1ada3a5203
ceff8cceee34f574d23909eb314c40c0
sframe_key: 3f7d9a7c83ae8e1c8a11ae695ab59314
b367e359fadac7b9c46b2bc6f81f46e1
6b96f0811868d59402b7e870102720b3
sframe_salt: 50b29329a04dc0f184ac3168
metadata: 4945544620534672616d65205747
nonce: 50b29329a04dc0f184ac740f
aad: 99012345674945544620534672616d65
205747
pt: 64726166742d696574662d736672616d
652d656e63
ct: 9901234567449408b6f490086165b9d6
f62b24ae1a59a56486b4ae8ed036b889
12e24f11Vector two, with cipher_suite 0x0002 🌸:
cipher_suite: 0x0002
kid: 0x0000000000000123
ctr: 0x0000000000004567
base_key: 000102030405060708090a0b0c0d0e0f
sframe_key_label: 534672616d6520312e30205365637265
74206b65792000000000000001230002
sframe_salt_label: 534672616d6520312e30205365637265
742073616c7420000000000000012300
02
sframe_secret: d926952ca8b7ec4a95941d1ada3a5203
ceff8cceee34f574d23909eb314c40c0
sframe_key: e2ec5c797540310483b16bf6e7a570d2
a27d192fe869c7ccd8584a8d9dab9154
9fbe553f5113461ec6aa83bf3865553e
sframe_salt: e68ac8dd3d02fbcd368c5577
metadata: 4945544620534672616d65205747
nonce: e68ac8dd3d02fbcd368c1010
aad: 99012345674945544620534672616d65
205747
pt: 64726166742d696574662d736672616d
652d656e63
ct: 99012345673f31438db4d09434e43afa
0f8a2f00867a2be085046a9f5cb4f101
d607Vector three, with cipher_suite 0x0003 🦋:
cipher_suite: 0x0003
kid: 0x0000000000000123
ctr: 0x0000000000004567
base_key: 000102030405060708090a0b0c0d0e0f
sframe_key_label: 534672616d6520312e30205365637265
74206b65792000000000000001230003
sframe_salt_label: 534672616d6520312e30205365637265
742073616c7420000000000000012300
03
sframe_secret: d926952ca8b7ec4a95941d1ada3a5203
ceff8cceee34f574d23909eb314c40c0
sframe_key: 2c5703089cbb8c583475e4fc461d97d1
8809df79b6d550f78eb6d50ffa80d892
11d57909934f46f5405e38cd583c69fe
sframe_salt: 38c16e4f5159700c00c7f350
metadata: 4945544620534672616d65205747
nonce: 38c16e4f5159700c00c7b637
aad: 99012345674945544620534672616d65
205747
pt: 64726166742d696574662d736672616d
652d656e63
ct: 990123456717fc8af28a5a695afcfc6c
8df6358a17e26b2fcb3bae32e443Vector four, with cipher_suite 0x0004 🧚:
cipher_suite: 0x0004
kid: 0x0000000000000123
ctr: 0x0000000000004567
base_key: 000102030405060708090a0b0c0d0e0f
sframe_key_label: 534672616d6520312e30205365637265
74206b65792000000000000001230004
sframe_salt_label: 534672616d6520312e30205365637265
742073616c7420000000000000012300
04
sframe_secret: d926952ca8b7ec4a95941d1ada3a5203
ceff8cceee34f574d23909eb314c40c0
sframe_key: d34f547f4ca4f9a7447006fe7fcbf768
sframe_salt: 75234edefe07819026751816
metadata: 4945544620534672616d65205747
nonce: 75234edefe07819026755d71
aad: 99012345674945544620534672616d65
205747
pt: 64726166742d696574662d736672616d
652d656e63
ct: 9901234567b7412c2513a1b66dbb4884
1bbaf17f598751176ad847681a69c6d0
b091c07018ce4adb34ebVector five, with cipher_suite 0x0005 👑:
cipher_suite: 0x0005
kid: 0x0000000000000123
ctr: 0x0000000000004567
base_key: 000102030405060708090a0b0c0d0e0f
sframe_key_label: 534672616d6520312e30205365637265
74206b65792000000000000001230005
sframe_salt_label: 534672616d6520312e30205365637265
742073616c7420000000000000012300
05
sframe_secret: 0fc3ea6de6aac97a35f194cf9bed94d4
b5230f1cb45a785c9fe5dce9c188938a
b6ba005bc4c0a19181599e9d1bcf7b74
aca48b60bf5e254e546d809313e083a3
sframe_key: d3e27b0d4a5ae9e55df01a70e6d4d28d
969b246e2936f4b7a5d9b494da6b9633
sframe_salt: 84991c167b8cd23c93708ec7
metadata: 4945544620534672616d65205747
nonce: 84991c167b8cd23c9370cba0
aad: 99012345674945544620534672616d65
205747
pt: 64726166742d696574662d736672616d
652d656e63
ct: 990123456794f509d36e9beacb0e261d
99c7d1e972f1fed787d4049f17ca2135
3c1cc24d56ceabced279Acknowledgements 🫶
The authors wish to specially thank Dr. Alex Gouaillard as one of the early contributors to the document. His passion and energy were key to the design and development of SFrame 🌈💖.
Contributors 💅
Frédéric Jacobs Apple Email: frederic.jacobs@apple.com
Marta Mularczyk Amazon Email: mulmarta@amazon.com
Suhas Nandakumar Cisco Email: snandaku@cisco.com
Tomas Rigaux Cisco Email: trigaux@cisco.com
Raphael Robert Phoenix R&D Email: ietf@raphaelrobert.com
Authors' Addresses 🦋
Emad Omara Apple Email: eomara@apple.com
Justin Uberti Fixie.ai Email: justin@fixie.ai
Sergio Garcia Murillo CoSMo Software Email: sergio.garcia.murillo@cosmosoftware.io
Richard Barnes (editor) Cisco Email: rlb@ipv.sx
Youenn Fablet Apple Email: youenn@apple.com
📦 where this page comes from
- ietf: rfc9605 2024-08
- 💖 girlified by claude-haiku-5-5 on : a language model's rewrite of the original, under its terms