RFC 8291

Message Encryption for Web Push

๐Ÿ“… published November 2017, from ietf

RFC 8291: Message Encryption for Web Push ๐ŸŽ€

Okay bestie, here's the deets! ๐Ÿ’– This is RFC 8291, called Message Encryption for Web Push, from the Internet Engineering Task Force (IETF). The author is M. Thomson at Mozilla. It's a Standards Track doc dated November 2017, with ISSN 2070-1721. The title block doesn't list anything it obsoletes or updates. โœจ

Abstract ๐ŸŒธ

So this doc is all about a message encryption scheme for Web Push ๐Ÿ” It keeps messages private and tamper-proof on their way from an application server to a user agent. It's giving confidentiality and integrity, no cap ๐Ÿซถ

Status of This Memo ๐Ÿฆ‹

Okay so this is an Internet Standards Track document. ๐Ÿ’…

It's a product of the IETF, which means it reflects the consensus of the IETF community. It went through public review and got approved for publication by the Internet Engineering Steering Group (IESG). If you want more info on Internet Standards, check out Section 2 of RFC 7841. ๐Ÿงš

For the current status of this document, any errata, and how to send feedback, head to https://www.rfc-editor.org/info/rfc8291. ๐Ÿ’œ

Copyright (c) 2017 IETF Trust and the persons identified as the document authors. All rights reserved. ๐ŸŽ€

This doc is subject to BCP 78 and the IETF Trust's Legal Provisions Relating to IETF Documents (https://trustee.ietf.org/license-info) as they were on the publication date. Please actually read these, bestie, because they explain your rights and restrictions for this doc ๐Ÿ™ˆ Any Code Components pulled out of this document have to include Simplified BSD License text, as described in Section 4.e of the Trust Legal Provisions, and they come with no warranty, as described in the Simplified BSD License. โญ

Table of Contents ๐Ÿ“

   1. Introduction ....................................................2
      1.1. Notational Conventions .....................................3
   2. Push Message Encryption Overview ................................3
      2.1. Key and Secret Distribution ................................4
   3. Push Message Encryption .........................................4
      3.1. Diffie-Hellman Key Agreement ...............................5
      3.2. Push Message Authentication ................................5
      3.3. Combining Shared and Authentication Secrets ................5
      3.4. Encryption Summary .........................................6
   4. Restrictions on Use of "aes128gcm" Content Coding ...............7
   5. Push Message Encryption Example .................................8
   6. IANA Considerations .............................................8
   7. Security Considerations .........................................8
   8. References .....................................................10
      8.1. Normative References ......................................10
      8.2. Informative References ....................................11
   Appendix A.  Intermediate Values for Encryption ...................12
   Author's Address ..................................................13

1. Introduction ๐Ÿงš

Okay so the Web Push protocol [RFC8030] is intermediated by necessity ๐Ÿ’… Messages from an application server get delivered to a user agent (UA) through a push service, as shown in Figure 1. It's a whole relay situation ๐Ÿฆ‹

    +-------+           +--------------+       +-------------+
    |  UA   |           | Push Service |       | Application |
    +-------+           +--------------+       +-------------+
        |                      |                      |
        |        Setup         |                      |
        |<====================>|                      |
        |           Provide Subscription              |
        |-------------------------------------------->|
        |                      |                      |
        :                      :                      :
        |                      |     Push Message     |
        |    Push Message      |<---------------------|
        |<---------------------|                      |
        |                      |                      |

                                 Figure 1

This doc explains how messages sent with this protocol can be kept safe from being inspected, modified, and forged by a push service ๐Ÿ”ฅ

Web Push messages are the payload of an HTTP message [RFC7230]. They get encrypted with an encrypted content encoding [RFC8188]. This doc covers how that content encoding is applied, plus a recommended key management scheme ๐Ÿซถ

Lowkey, lots of Web Push users on the same user agent share one central agent that bundles up push functionality. That agent can enforce this encryption scheme on apps that use push messaging โœจ An agent that only delivers properly encrypted messages is strongly encouraging end-to-end protection, and that's iconic ๐Ÿ‘‘

A web browser that implements the Push API [API] can enforce encryption too, by only forwarding messages that were properly encrypted ๐Ÿ’–

1.1. Notational Conventions ๐ŸŽ€

The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "NOT RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in BCP 14 [RFC2119] [RFC8174] when, and only when, they appear in all capitals, as shown here. That's the rule, bestie, the caps are the whole point ๐Ÿ”

This doc uses the terminology from [RFC8030], mainly "user agent", "push service", and "application server" ๐Ÿ’œ

2. Push Message Encryption Overview ๐ŸŒˆ

Encrypting a push message uses Elliptic Curve Diffie-Hellman (ECDH) [ECDH] on the P-256 curve [FIPS186]. That gets you a shared secret (see Section 3.1) and a symmetric secret for authentication (see Section 3.2) ๐Ÿฆ‹

A user agent makes an ECDH key pair and an authentication secret for each subscription it creates. It sends the ECDH public key and the authentication secret to the application server along with the other details of the push subscription ๐Ÿ“

When sending a message, an application server makes an ECDH key pair and a random salt. The ECDH public key goes into the "keyid" parameter of the encrypted content coding header, and the salt goes into the "salt" parameter of that same header (see Section 2.1 of [RFC8188]). The ECDH key pair can be tossed once the message is encrypted ๐Ÿงš

The message content gets encrypted or decrypted with a content encryption key and nonce. These values are derived by taking the "keyid" and "salt" as input to the process described in Section 3 โœจ

2.1. Key and Secret Distribution ๐Ÿซถ

The application using the subscription hands the subscription public key and authentication secret to an authorized application server. This can ride along with other subscription info from the user agent, like the push subscription URI ๐Ÿ’…

An application MUST use an authenticated, confidentiality-protected communications medium for this. Not optional, bestie. On top of the reasons in [RFC8030], this keeps the authentication secret away from unauthorized entities. If they got it, they could generate push messages that the user agent would accept ๐Ÿ˜ญ

Most apps that use push messaging already have a relationship with an application server that can handle subscription data distribution ๐ŸŒธ An authenticated mechanism with enough confidentiality and integrity protection, like HTTPS [RFC2818], is good enough ๐ŸŽ€

3. Push Message Encryption ๐Ÿ’–

Push message encryption happens in four phases:

  • A shared secret is derived using ECDH [ECDH] (see Section 3.1 of this document).
  • The shared secret is then combined with the authentication secret to produce the input keying material (IKM) used in [RFC8188] (see Section 3.3 of this document).
  • A content encryption key and nonce are derived using the process in [RFC8188].
  • Encryption or decryption follows according to [RFC8188].

The key derivation process is summarized in Section 3.4. Restrictions on the use of the encrypted content coding are described in Section 4 ๐Ÿ”ฅ

3.1. Diffie-Hellman Key Agreement ๐ŸŒˆ

For each new subscription the user agent makes for an app, it also makes a P-256 [FIPS186] key pair for ECDH [ECDH] ๐Ÿฆ‹

When sending a push message, the application server also makes a new ECDH key pair on that same P-256 curve ๐Ÿ’œ

The application server's ECDH public key is included as the "keyid" parameter in the encrypted content coding header (see Section 2.1 of [RFC8188]) ๐Ÿ“

The application server combines its ECDH private key with the public key from the user agent using the process in [ECDH]. When the push message arrives, the user agent combines its private key with the public key the application server sent in the "keyid" parameter, the same way. Both sides end up with the same ECDH shared secret, which is iconic ๐Ÿ‘‘

3.2. Push Message Authentication ๐Ÿงš

To make sure push messages get authenticated correctly, a symmetric authentication secret gets added to the info the user agent generates. That authentication secret gets mixed into the key derivation process in Section 3.3 โœจ

A user agent MUST generate and provide a hard-to-guess sequence of 16 octets for authenticating push messages. It SHOULD be generated by a cryptographically strong random number generator [RFC4086]. Translation: make it unguessable, bestie ๐Ÿ”

3.3. Combining Shared and Authentication Secrets ๐Ÿ’…

The shared secret from ECDH gets combined with the authentication secret using the HMAC-based key derivation function (HKDF) [RFC5869]. The result is the input keying material used by [RFC8188] ๐Ÿซถ

The HKDF function uses the SHA-256 hash algorithm [FIPS180-4] with these inputs:

  • salt: the authentication secret
  • IKM: the shared secret derived using ECDH
  • info: the concatenation of the ASCII-encoded string "WebPush: info" (this string is not NUL-terminated), a zero octet, the user agent ECDH public key, and the application server ECDH public key, (both ECDH public keys are in the uncompressed point form defined in [X9.62]. That is:
key_info = "WebPush: info" || 0x00 || ua_public || as_public
  • L: 32 octets (i.e., the output is the length of the underlying SHA-256 HMAC function output)

3.4. Encryption Summary โœจ

okay so this is where it all comes together, bestie ๐Ÿ’… we end up with one final content encryption key and nonce. Here's the whole sequence in pseudocode, with HKDF split into separate steps and run through HMAC with SHA-256 ๐Ÿ”

   -- For a user agent:
   ecdh_secret = ECDH(ua_private, as_public)
   auth_secret = random(16)
   salt = <from content coding header>

   -- For an application server:
   ecdh_secret = ECDH(as_private, ua_public)
   auth_secret = <from user agent>
   salt = random(16)

   -- For both:

   ## Use HKDF to combine the ECDH and authentication secrets
   # HKDF-Extract(salt=auth_secret, IKM=ecdh_secret)
   PRK_key = HMAC-SHA-256(auth_secret, ecdh_secret)
   # HKDF-Expand(PRK_key, key_info, L_key=32)
   key_info = "WebPush: info" || 0x00 || ua_public || as_public
   IKM = HMAC-SHA-256(PRK_key, key_info || 0x01)

   ## HKDF calculations from RFC 8188
   # HKDF-Extract(salt, IKM)
   PRK = HMAC-SHA-256(salt, IKM)
   # HKDF-Expand(PRK, cek_info, L_cek=16)
   cek_info = "Content-Encoding: aes128gcm" || 0x00
   CEK = HMAC-SHA-256(PRK, cek_info || 0x01)[0..15]
   # HKDF-Expand(PRK, nonce_info, L_nonce=12)
   nonce_info = "Content-Encoding: nonce" || 0x00
   NONCE = HMAC-SHA-256(PRK, nonce_info || 0x01)[0..11]

Quick heads-up: this skips the exclusive-OR of the final nonce with the record sequence number. Why? Push messages only ever have a single record (see Section 4), and the sequence number of that first record is zero, so there's nothing to XOR ๐Ÿซถ


4. Restrictions on Use of "aes128gcm" Content Coding ๐ŸŽ€

okay this section is basically the rules of the group chat, and they're strict ๐Ÿ˜ค

An application server MUST encrypt a push message with a single record. Why? It lets the receiver get away with a super minimal implementation that only has to handle one record ๐Ÿฆ‹ The application server MUST also set the "rs" parameter in the "aes128gcm" content coding header to a size that is greater than the sum of the lengths of the plaintext, the padding delimiter (1 octet), any padding, and the authentication tag (16 octets). No cutting corners on that math, periodt ๐Ÿ’…

A push message MUST include the application server ECDH public key in the "keyid" parameter of the encrypted content coding header. The uncompressed point form defined in [X9.62] (that is, a 65-octet sequence that starts with a 0x04 octet) forms the entirety of the "keyid". Lowkey unexpected, but this means the "keyid" parameter will not be valid UTF-8 as recommended in [RFC8188] ๐ŸŒธ

A push service is not required to support more than 4096 octets of payload body (see Section 7.2 of [RFC8030]). Once you take away the header (86 octets), the padding (minimum 1 octet), and the expansion for AEAD_AES_128_GCM (16 octets), you're left with, at most, 3993 octets of plaintext ๐Ÿงฎ

An application server MUST NOT use other content encodings for push messages. Compression is the big no-no here, because content encodings that compress could result in leaking of push message contents ๐Ÿ™ˆ So the Content-Encoding header field has exactly one value, which is "aes128gcm". Multiple "aes128gcm" values are not permitted. Not even a little bit ๐Ÿ’€

A user agent is not required to support multiple records. A user agent MAY ignore the "rs" parameter. If a record size is unchecked, decryption will fail with high probability for all valid cases. The padding delimiter octet MUST be checked; values other than 0x02 MUST cause the message to be discarded. Basically, if the delimiter isn't 0x02, it's getting tossed ๐Ÿ—‘๏ธ

Real talk: the padding delimiter check is non-negotiable. Skip it and your receiver is letting in stuff it shouldn't ๐Ÿ˜ญ

5. Push Message Encryption Example ๐Ÿงš

okay, time for the fun part, a real example! Here's a push message getting sent to a push service ๐Ÿ“

POST /push/JzLQ3raZJfFBR0aqvOMsLrt54w4rJUsV HTTP/1.1
Host: push.example.net
TTL: 10
Content-Length: 145
Content-Encoding: aes128gcm

DGv6ra1nlYgDCS1FRnbzlwAAEABBBP4z9KsN6nGRTbVYI_c7VJSPQTBtkgcy27ml
mlMoZIIgDll6e3vCYLocInmYWAmS6TlzAC8wEqKK6PBru3jl7A_yl95bQpu6cVPT
pK4Mqgkf1CXztLVBSt2Ks3oZwbuwXPXLWyouBWLVWGNWQexSgSxsj_Qulcy4a-fN

The ASCII-encoded string in there is "When I grow up, I want to be a watermelon" ๐Ÿ‰ The content body is shown with line wrapping and URL-safe base64url [RFC4648] encoding so it fits on the page ๐ŸŒˆ

Here are the keys, using the uncompressed form [X9.62] encoded with base64url โœจ

   Authentication Secret: BTBZMqHH6r4Tts7J_aSIgg
   Receiver:
      private key: q1dXpw3UpT5VOmu_cf_v6ih07Aems3njxI-JWgLcM94
      public key: BCVxsr7N_eNgVRqvHtD0zTZsEc6-VV-JvLexhqUzORcx
                  aOzi6-AYWXvTBHm4bjyPjs7Vd8pZGH6SRpkNtoIAiw4
   Sender:
      private key: yfWPiYE-n46HLnH0KqZOF1fJJU3MYrct3AELtAQ-oRw
      public key: BP4z9KsN6nGRTbVYI_c7VJSPQTBtkgcy27mlmlMoZIIg
                  Dll6e3vCYLocInmYWAmS6TlzAC8wEqKK6PBru3jl7A8

If you want the intermediate values, they're all in Appendix A ๐Ÿ”


6. IANA Considerations ๐ŸŒท

This document does not require any IANA actions. Nothing to register, no forms to file, we're literally chilling ๐Ÿ˜Œ


7. Security Considerations ๐Ÿ”

okay this is the serious bit, so pay attention bestie ๐Ÿ‘‘ Everything here matters, no skipping ๐Ÿ˜ญ

The privacy and security considerations of [RFC8030] all apply to the use of this mechanism. So all of that stuff applies here too ๐Ÿซก

The Security Considerations section of [RFC8188] describes the limitations of the content encoding. The big one: no HTTP header fields are protected by the content encoding scheme. So a user agent MUST consider HTTP header fields to have come from the push service. Trust the header fields as coming from the push service, not from the app server ๐Ÿ‘€

Though header fields might be necessary for processing an HTTP response correctly, they are not needed for correct operation of the protocol. An application on the user agent that uses information from header fields to alter their processing of a push message is exposed to a risk of attack by the push service. So be careful if your app is doing that, it's a whole vulnerability ๐Ÿ’€

The timing and length of communication cannot be hidden from the push service. An outside observer might see individual messages intermixed with each other, but the push service will see which application server is talking to which user agent and the subscription that is used. Also, the length of messages could be revealed unless the padding provided by the content encoding scheme is used to obscure length. So use that padding ๐Ÿ™

Public key validation is a MUST-level vibe check, and it's a big deal. Failing to validate a public key can allow an attacker to extract a private key. Yikes ๐Ÿ˜ฑ

The user agent and application MUST verify that the public key they receive is on the P-256 curve. The appropriate validation procedures are defined in Section 4.3.7 of [X9.62] and, alternatively, in Section 5.6.2.3 of [KEYAGREEMENT]. This process consists of three steps:

  1. Verify that Y is not the point at infinity (O),
  2. Verify that for Y = (x, y), both integers are in the correct interval,
  3. Ensure that (x, y) is a correct solution to the elliptic curve equation.

For these curves, implementers do not need to verify membership in the correct subgroup. Less work for you, bestie ๐Ÿฆ‹

If this encryption scheme ever needs replacing, a new content coding scheme could be defined. To manage progressive deployment of the new scheme, the user agent can expose information on the content coding schemes that it supports. The supportedContentEncodings parameter of the Push API [API] is an example of how this might be done ๐ŸŒˆ


8. References ๐Ÿ“š

8.1. Normative References

8.2. Informative References

Appendix A. Intermediate Values for Encryption โœจ

okay bestie, this is the behind-the-scenes closet ๐Ÿงš every intermediate value from the Section 5 example, all laid out so you can check the math yourself. the base64url values have whitespace in them that you can totally remove ๐Ÿ’…

First, the inputs to the calculation:

Plaintext:  V2hlbiBJIGdyb3cgdXAsIEkgd2FudCB0byBiZSBhIHdhdGVybWVsb24

Application server public key (as_public):
   BP4z9KsN6nGRTbVYI_c7VJSPQTBtkgcy27mlmlMoZIIg
   Dll6e3vCYLocInmYWAmS6TlzAC8wEqKK6PBru3jl7A8

Application server private key (as_private):
   yfWPiYE-n46HLnH0KqZOF1fJJU3MYrct3AELtAQ-oRw

User agent public key (ua_public):  BCVxsr7N_eNgVRqvHtD0zTZsEc6-VV-
   JvLexhqUzORcx aOzi6-AYWXvTBHm4bjyPjs7Vd8pZGH6SRpkNtoIAiw4

User agent private key (ua_private):
   q1dXpw3UpT5VOmu_cf_v6ih07Aems3njxI-JWgLcM94

Salt:  DGv6ra1nlYgDCS1FRnbzlw

Authentication secret (auth_secret):  BTBZMqHH6r4Tts7J_aSIgg

Note that knowledge of just one of the private keys is necessary. ๐Ÿ”‘ The application server randomly generates the salt value, whereas salt is input to the receiver. ๐Ÿฅบ

Now for the intermediate values this produces:

Shared ECDH secret (ecdh_secret):
   kyrL1jIIOHEzg3sM2ZWRHDRB62YACZhhSlknJ672kSs

Pseudorandom key (PRK) for key combining (PRK_key):
   Snr3JMxaHVDXHWJn5wdC52WjpCtd2EIEGBykDcZW32k

Info for key combining (key_info):  V2ViUHVzaDogaW5mbwAEJXGyvs3942BVG
   q8e0PTNNmwR zr5VX4m8t7GGpTM5FzFo7OLr4BhZe9MEebhuPI-OztV3
   ylkYfpJGmQ22ggCLDgT-M_SrDepxkU21WCP3O1SUj0Ew
   bZIHMtu5pZpTKGSCIA5Zent7wmC6HCJ5mFgJkuk5cwAv MBKiiujwa7t45ewP

Input keying material for content encryption key derivation (IKM):
   S4lYMb_L0FxCeq0WhDx813KgSYqU26kOyzWUdsXYyrg

PRK for content encryption (PRK):
   09_eUZGrsvxChDCGRCdkLiDXrReGOEVeSCdCcPBSJSc

Info for content encryption key derivation (cek_info):
   Q29udGVudC1FbmNvZGluZzogYWVzMTI4Z2NtAA

Content encryption key (CEK):  oIhVW04MRdy2XN9CiKLxTg

Info for content encryption nonce derivation (nonce_info):
   Q29udGVudC1FbmNvZGluZzogbm9uY2UA

Nonce (NONCE):  4h_95klXJ5E_qnoN

The salt, record size of 4096, and application server public key produce an 86-octet header of: ๐ŸŽ€

DGv6ra1nlYgDCS1FRnbzlwAAEABBBP4z 9KsN6nGRTbVYI_c7VJSPQTBtkgcy27ml
mlMoZIIgDll6e3vCYLocInmYWAmS6Tlz AC8wEqKK6PBru3jl7A8

Then the push message plaintext gets the padding delimiter octet (0x02) appended, so it becomes: ๐Ÿ“

V2hlbiBJIGdyb3cgdXAsIEkgd2FudCB0 byBiZSBhIHdhdGVybWVsb24C

The plaintext is then encrypted with AES-GCM, which emits this ciphertext: ๐Ÿ’–

8pfeW0KbunFT06SuDKoJH9Ql87S1QUrd irN6GcG7sFz1y1sqLgVi1VhjVkHsUoEs
bI_0LpXMuGvnzQ

And finally, the header and ciphertext get concatenated and produce the result shown in Section 5. That's the whole glow-up, periodt! ๐Ÿฆ‹โœจ

Author's Address ๐Ÿซถ

Martin Thomson Mozilla

Email: martin.thomson@gmail.com

๐Ÿ“ฆ where this page comes from