Skip to content

Latest commit

 

History

History
executable file
·
1773 lines (1367 loc) · 67.3 KB

draft-ietf-mls-extensions.md

File metadata and controls

executable file
·
1773 lines (1367 loc) · 67.3 KB

title: The Messaging Layer Security (MLS) Extensions abbrev: MLS docname: draft-ietf-mls-extensions-latest submissiontype: IETF category: std

number: date: consensus: true v: 3 area: "Security" workgroup: "Messaging Layer Security" keyword:

author:

contributor:

normative:

informative:

--- abstract

The Messaging Layer Security (MLS) protocol is an asynchronous group authenticated key exchange protocol. MLS provides a number of capabilities to applications, as well as several extension points internal to the protocol. This document provides a consolidated application API, guidance for how the protocol's extension points should be used, and a few concrete examples of both core protocol extensions and uses of the application API.

--- middle

Introduction

This document defines extensions to MLS {{!RFC9420}} that are not part of the main protocol specification, and uses them to explain how to extend the core operation of the MLS protocol. It also describes how applications can safely interact with the MLS to take advantage of security features of MLS.

The MLS protocol is designed to be integrated into applications, in order to provide security services that the application requires. There are two questions to answer when designing such an integration:

  1. How does the application provide the services that MLS requires?
  2. How does the application use MLS to get security benefits?

The MLS Architecture {{?I-D.ietf-mls-architecture}} describes the requirements for the first of these questions, namely the structure of the Delivery Service and Authentication Service that MLS requires. The next section of this document focuses on the second question.

MLS itself offers some basic functions that applications can use, such as the secure message encapsulation (PrivateMessage), the MLS exporter, and the epoch authenticator. Current MLS applications make use of these mechanisms to acheive a variety of confidentiality and authentication properties.

As application designers become familiar with MLS, there is interest in leveraging other cryptographic tools that an MLS group provides:

  • HPKE (Hybrid Public Key Encryption {{!RFC9180}}) and signature key pairs for each member, where the private key is known only to that member, and the public key is authenticated to the other members.

  • A pre-shared key mechanism that can allow an application to inject data into the MLS key schedule.

  • An exporter mechanism that allows applications to derive secrets from the MLS key schedule.

  • Association of data with Commits as a synchronization mechanism.

  • Binding of information to the GroupContext to confirm group agreement.

There is also interest in exposing an MLS group to multiple loosely-coordinated components of an application. To accommodate such cases, the above mechanisms need to be exposed in such a way that the usage of different components do not conflict with each other, or with MLS itself.

This document defines a set of mechanisms that application components can use to ensure that their use of these facilities is properly domain-separated from MLS itself, and from other application components that might be using the same MLS group.

Conventions and Definitions

{::boilerplate bcp14-tagged}

This document makes heavy use of the terminology and the names of structs in the MLS specification {{!RFC9420}}. In addition, we introduce the following new terms:

Application: : The system that instantiates, manages, and uses an MLS group. Each MLS group is used by exactly one application, but an application may maintain multiple groups.

Application component: : A subsystem of an application that has access to an MLS group.

Component ID: : An identifier for an application component. These identifiers are assigned by the application.

Developing Extensions for the MLS Protocol

MLS is highly extensible, and was designed to be used in a variety of different applications. As such, it is important to separate extensions that change the behavior of an MLS stack, and application usages of MLS that just take advantage of features of MLS or specific extensions (hereafter referred to as components). Furthermore it is essential that application components do not change the security properties of MLS or require new security analysis of the MLS protocol itself.

The Safe Application Interface

The mechansms in this section take MLS mechanisms that are either not inherently designed to be used by applications, or not inherently designed to be used by multiple application components, and adds a domain separator that separates application usage from MLS usage, and application components' usage from each other:

  • Public-key encryption operations are tagged so that encrypted data will only decrypt in the context of a given component.

  • Signing operations are similarly tagged so that signatures will only verify in the context of a given component.

  • Exported values include an identifier for the component to which they are being exported, so that different components will get different exported values.

  • Pre-shared keys are identified as originating from a specific component, so that differnet components' contributions to the MLS key schedule will not collide.

  • Additional Authenticated Data (AAD) can be domain separated by component.

Similarly, the content of application messages (application_data) can be distinguished and routed to different parts of an application according to the media type of that content using the content negotiation mechanism defined in {{content-advertisement}}.

We also define new general mechanisms that allow applications to take advantage of the extensibility mechanisms of MLS without having to define extensions themselves:

  • An app_data_dictionary extension type that associates application data with MLS messages, or with the state of the group.

  • An AppEphemeral proposal type that enables arbitrary application data to be associated to a Commit.

  • An AppDataUpdate proposal type that enables efficient updates to an app_data_dictionary GroupContext extension.

As with the above, information carried in these proposals and extension marked as belonging to a specific application component, so that components can manage their information independently.

The separation between components is acheived by the application assigning each component a unique component ID number. These numbers are then incorporated into the appopriate calculations in the protocol to achieve the required separation.

Component IDs

A component ID is a four-byte value that uniquely identifies a component within the scope of an application.

uint32 ComponentID;

TODO: What are the uniqueness requirements on these? It seems like the more diversity, the better. For example, if a ComponentID is reused across applications (e.g., via an IANA registry), then there will be a risk of replay across applications. Maybe we should include a binder to the group/epoch as well, something derived from the key schedule.

TODO: It might be better to frame these in terms of "data types" instead of components, to avoid presuming software architecture. Though that makes less sense for the more "active" portions of the API, e.g., signing and encryption.

When a label is required for an operation, the following data structure is used. The label field identifies the operation being performed. The component_id field identifies the component performing the operation. The context field is specified by the operation in question.

struct {
  opaque label<V>;
  ComponentID component_id;
  opaque context<V>;
} ComponentOperationLabel;

Hybrid Public Key Encryption (HPKE) Keys {#safe-hpke}

This component of the API allows components to make use of the HPKE key pairs generated by MLS. A component identified by a ComponentID can use any HPKE key pair for any operation defined in {{!RFC9180}}, such as encryption, exporting keys and the PSK mode, as long as the info input to Setup<MODE>S and Setup<MODE>R is set to ComponentOperationLabel with component_id set to the appopriate ComponentID. The context can be set to an arbitrary Context specified by the application designer and can be empty if not needed. For example, a component can use a key pair PublicKey, PrivateKey to encrypt data as follows:

SafeEncryptWithContext(ComponentID, PublicKey, Context, Plaintext) =
  SealBase(PublicKey, ComponentOperationLabel, "", Plaintext)

SafeDecryptWithContext(ComponentID, PrivateKey, Context,
  KEMOutput, Ciphertext) = OpenBase(KEMOutput, PrivateKey,
                             ComponentOperationLabel, "", Ciphertext)

Where the fields of ComponentOperationLabel are set to

label = "MLS 1.0 Application"
component_id = ComponentID
context = Context

TODO: Should this use EncryptWithLabel / DecryptWithLabel? That wouldn't cover other modes / exports, but you could say "mutatis mutandis".

For operations involving the secret key, ComponentID MUST be set to the ComponentID of the component performing the operation, and not to the ID of any other component. In particular, this means that a component cannot decrypt data meant for another component, while components can encrypt data that other components can decrypt.

In general, a ciphertext encrypted with a PublicKey can be decrypted by any entity who has the corresponding PrivateKey at a given point in time according to the MLS protocol (or application component). For convenience, the following list summarizes lifetimes of MLS key pairs.

  • The key pair of a non-blank ratchet tree node. The PrivateKey of such a key pair is known to all members in the node’s subtree. In particular, a PrivateKey of a leaf node is known only to the member in that leaf. A member in the subtree stores the PrivateKey for a number of epochs, as long as the PublicKey does not change. The key pair of the root node SHOULD NOT be used, since the external key pair recalled below gives better security.
  • The external_priv, external_pub key pair used for external initialization. The external_priv key is known to all group members in the current epoch. A member stores external_priv only for the current epoch. Using this key pair gives better security guarantees than using the key pair of the root of the ratchet tree and should always be preferred.
  • The init_key in a KeyPackage and the corresponding secret key. The secret key is known only to the owner of the KeyPackage and is deleted immediately after it is used to join a group.

Signature Keys

MLS session states contain a number of signature keys including the ones in the LeafNode structs. Application components can safely sign content and verify signatures using these keys via the SafeSignWithLabel and SafeVerifyWithLabel functions, respectively, much like how the basic MLS protocol uses SignWithLabel and VerifyWithLabel.

In more detail, a component identified by ComponentID should sign and verify using:

SafeSignWithLabel(ComponentID, SignatureKey, Label, Content) =
   SignWithLabel(SignatureKey, "ComponentOperationLabel",
      ComponentOperationLabel)

SafeVerifyWithLabel(ComponentID, VerificationKey, Label, Content,
  SignatureValue) = VerifyWithLabel(VerificationKey,
                      "ComponentOperationLabel",
                       ComponentOperationLabel,
                       SignatureValue)

Where the fields of ComponentOperationLabel are set to

label = Label
component_id = ComponentID
context = Content

For signing operations, the ComponentID MUST be set to the ComponentID of the component performing the signature, and not to the ID of any other component. This means that a component cannot produce signatures in place of other component. However, components can verify signatures computed by other components. Domain separation is ensured by explicitly including the ComponentID with every operation.

Exported Secrets

An application component can use MLS as a group key agreement protocol by exporting symmetric keys. Such keys can be exported (i.e. derived from MLS key material) in two phases per epoch: Either at the start of the epoch, or during the epoch. Derivation at the start of the epoch has the added advantage that the source key material is deleted after use, allowing the derived key material to be deleted later even during the same MLS epoch to achieve forward secrecy. The following protocol secrets can be used to derive key from for use by application components:

  • exporter_secret at the beginning of an epoch
  • application_export_secret during an epoch

The application_export_secret is an additional secret derived from the epoch_secret at the beginning of the epoch in the same way as the other secrets listed in Table 4 of {{!RFC9420}} using the label "application_export".

Any derivation performed by an application component either from the exporter_secret or the application_export_secret has to use the following function:

DeriveApplicationSecret(Secret, Label) =
  ExpandWithLabel(Secret, "ApplicationExport " +
                  ComponentID + " " + Label)

Where ExpandWithLabel is defined in {{Section 8 of RFC9420}} and where ComponentID MUST be set to the ComponentID of the component performing the export.

TODO: This section seems over-complicated to me. Why is it not sufficient to just use the exporter_secret? Or the MLS-Exporter mechanism with a label structured to include the ComponentID?

Pre-Shared Keys (PSKs)

PSKs represent key material that is injected into the MLS key schedule when creating or processing a commit as defined in {{Section 8.4 of !RFC9420}}. Its injection into the key schedule means that all group members have to agree on the value of the PSK.

While PSKs are typically cryptographic keys which due to their properties add to the overall security of the group, the PSK mechanism can also be used to ensure that all members of a group agree on arbitrary pieces of data represented as octet strings (without the necessity of sending the data itself over the wire). For example, a component can use the PSK mechanism to enforce that all group members have access to and agree on a password or a shared file.

This is achieved by creating a new epoch via a PSK proposal. Transitioning to the new epoch requires using the information agreed upon.

To facilitate using PSKs in a safe way, this document defines a new PSKType for application components. This provides domain separation between pre-shared keys used by the core MLS protocol and applications, and between those used by different components.

enum {
  // ...
  application(3),
  (255)
} PSKType;

struct {
  PSKType psktype;
  select (PreSharedKeyID.psktype) {
    // ...
    case application:
      ComponentID component_id;
      opaque psk_id<V>;
  };
  opaque psk_nonce<V>;
} PreSharedKeyID;

TODO: It seems like you could also do this by structuring the external PSKType as (component_id, psk_id). I guess this approach separates this API from other external PSKs.

Attaching Application Data to MLS Messages

The MLS GroupContext, LeafNode, KeyPackage, and GroupInfo objects each have an extensions field that can carry additional data not defined by the MLS specification. The app_data_dictionary extension provides a generic container that applications can use to attach application data to these messages. Each usage of the extension serves a slightly different purpose:

  • GroupContext: Confirms that all members of the group agree on the application data, and automatically distributes it to new joiners.

  • KeyPackage and LeafNode: Associates the application data to a particular client, and advertises it to the other members of the group.

  • GroupInfo: Distributes the application data confidentially to the new joiners for whom the GroupInfo is encrypted (as a Welcome message).

The content of the app_data_dictionary extension is a serialized AppDataDictionary object:

struct {
    ComponentID component_id;
    opaque data<V>;
} ComponentData;

struct {
    ComponentData component_data<V>;
} AppDataDictionary;

The entries in the component_data MUST be sorted by component_id, and there MUST be at most one entry for each component_id.

An app_data_dictionary extension in a LeafNode, KeyPackage, or GroupInfo can be set when the object is created. An app_data_dictionary extension in the GroupContext needs to be managed using the tools available to update GroupContext extensions. The creator of the group can set extensions unilaterally. Thereafter, the AppDataUpdate proposal described in the next section is used to update the app_data_dictionary extension.

Updating Application Data in the GroupContext {#appdataupdate}

Updating the app_data_dictionary with a GroupContextExtensions proposal is cumbersome. The application data needs to be transmitted in its entirety, along with any other extensions, whether or not they are being changed. And a GroupContextExtensions proposal always requires an UpdatePath, which updating application state never should.

The AppDataUpdate proposal allows the app_data_dictionary extension to be updated without these costs. Instead of sending the whole value of the extension, it sends only an update, which is interpreted by the application to provide the new content for the app_data_dictionary extension. No other extensions are sent or updated, and no UpdatePath is required.

enum {
    invalid(0),
    update(1),
    remove(2),
    (255)
} AppDataUpdateOperation;

struct {
    ComponentID component_id;
    AppDataUpdateOperation op;

    select (AppDataUpdate.op) {
        case update: opaque update<V>;
        case remove: struct{};
    };
} AppDataUpdate;

An AppDataUpdate proposal is invalid if its component_id references a component that is not known to the application, or if it specifies the removal of state for a component_id that has no state present. A proposal list is invalid if it includes multiple AppDataUpdate proposals that remove state for the same component_id, or proposals that both update and remove state for the same component_id. In other words, for a given component_id, a proposal list is valid only if it contains (a) a single remove operation or (b) one or more update operation.

AppDataUpdate proposals are processed after any default proposals (i.e., those defined in {{RFC9420}}), and any AppEphemeral proposals (defined in {{app-ephemeral}}).

When an MLS group contains the AppDataUpdate proposal type in the proposal_types list in the group's required_capabilities extension, a GroupContextExtensions proposal MUST NOT add, remove, or modify the app_data_dictionary GroupContext extension. In other words, when every member of the group supports the AppDataUpdate proposal, a GroupContextExtensions proposal could be sent to update some other extension(s), but the app_data_dictionary GroupContext extension, if it exists, is left as it was.

A commit can contain a GroupContextExtensions proposal which modifies GroupContext extensions other than app_data_dictionary, and can be followed by zero or more AppDataUpdate proposals. This allows modifications to both the app_data_dictionary extension (via AppDataUpdate) and other extensions (via GroupContextExtensions) in the same Commit.

A client applies AppDataUpdate proposals by component ID. For each component_id field that appears in an AppDataUpdate proposal in the Commit, the client assembles a list of AppDataUpdate proposals with that component_id, in the order in which they appear in the Commit, and processes them in the following way:

  • If the list comprises a single proposal with the op field set to remove:

    • If there is an entry in the component_states vector in the application_state extension with the specified component_id, remove it.

    • Otherwise, the proposal is invalid.

  • If the list comprises one or more proposals, all with op field set to update:

    • Provide the application logic registered to the component_id value with the content of the update field from each proposal, in the order specified.

    • The application logic returns either an opaque value new_data that will be stored as the new application data for this component, or else an indication that it considers this update invalid.

    • If the application logic considers the update invalid, the MLS client MUST consider the proposal list invalid.

    • If no app_data_dictionary extension is present in the GroupContext, add one to the end of the extensions list in the GroupContext.

    • If there is an entry in the component_data vector in the app_data_dictionary extension with the specified component_id, then set its data field to the specified new_data.

    • Otherwise, insert a new entry in the component_states vector with the specified component_id and the data field set to the new_data value. The new entry is inserted at the proper point to keep the component_states vector sorted by component_id.

  • Otherwise, the proposal list is invalid.

NOTE: An alternative design here would be to have the update operation simply set the new value for the app_data_dictionary GCE, instead of sending a diff. This would be simpler in that the MLS stack wouldn't have to ask the application for the new state value, and would discourage applications from storing large state in the GroupContext directly (which bloats Welcome messages). It would effectively require the state in the GroupContext to be a hash of the real state, to avoid large AppDataUpdate proposals. This pushes some complexity onto the application, since the application has to define a hashing algorithm, and define its own scheme for initializing new joiners.

AppDataUpdate proposals do not require an UpdatePath. An AppDataUpdate proposal can be sent by an external sender. Likewise, AppDataUpdate proposals can be included in an external commit. Applications can make more restrictive validity rules for the update of their components, such that some components would not be valid at the application when sent in an external commit or via an external proposer.

Attaching Application Data to a Commit {#app-ephemeral}

The AppEphemeral proposal type allows an application component to associate application data to a Commit, so that the member processing the Commit knows that all other group members will be processing the same data. AppEphemeral proposals are ephemeral in the sense that they do not change any persistent state related to MLS, aside from their appearance in the transcript hash.

The content of an AppEphemeral proposal is the same as an app_data_dictionary extension. The proposal type is set in {{iana-considerations}}.

struct {
    ComponentID component_id;
    opaque data<V>;
} AppEphemeral;

An AppEphemeral proposal is invalid if it contains a component_id that is unknown to the application, or if the app_data_dictionary field contains any ComponentData entry whose data field is considered invalid by the application logic registered to the indicated component_id.

AppEphemeral proposals MUST be processed after any default proposals (i.e., those defined in {{RFC9420}}), but before any AppDataUpdate proposals.

A client applies an AppEphemeral proposal by providing the contents of the app_data_dictionary field to the component identified by the component_id. If a Commit references more than one AppEphemeral proposal for the same component_id value, then they MUST be processed in the order in which they are specified in the Commit.

AppEphemeral proposals do not require an UpdatePath. An AppEphemeral proposal can be sent by an external sender. Likewise, AppEphemeral proposals can be included in an external commit. Applications can make more restrictive validity rules for ephemeral updates of their components, such that some components would not be valid at the application when sent in an external commit or via an external proposer.

Safe Additional Authenticated Data (AAD) {#safe-aad}

The PrivateContentAAD struct in MLS can contain arbitrary additional application-specific AAD in its authenticated_data field. This API defines a framing used to allow multiple extensions to add AAD safely without conflicts or ambiguity.

When any AAD safe extension is included in the authenticated_data field, the "safe" AAD items MUST come before any non-safe data in the authenticated_data field. Safe AAD items are framed using the SafeAAD struct and are sorted in increasing numerical order of the component_id. The struct is described below:

struct {
  ComponentID component_id;
  opaque aad_item_data<V>;
} SafeAADItem;

struct {
  SafeAADItem aad_items<V>;
} SafeAAD;

If the SafeAAD is present or not in the authenticated_data is determined by the presence of the safe_aad component in the app_data_dictionary extension in the GroupContext (see {{negotiation}}). If safe_aad is present, but none of the "safe" AAD components have data to send in a particular message, the aad_items is a zero-length vector.

Negotiating Extensions and Components {#negotiation}

MLS defines a Capabilities struct for LeafNodes (in turn used in KeyPackages), which describes which extensions are supported by the associated node. However, that struct (defined in {{Section 7.2 of !RFC9420}}) only has fields for a subset of the extensions possible in MLS, as reproduced below.

struct {
    ProtocolVersion versions<V>;
    CipherSuite cipher_suites<V>;
    ExtensionType extensions<V>;
    ProposalType proposals<V>;
    CredentialType credentials<V>;
} Capabilities;

The "MLS Extensions Types" registry represents extensibility of four core structs (GroupContext, GroupInfo, KeyPackage, and LeafNode) that have far reaching effects on the use of the protocol. The majority of MLS extensions in {{!RFC9420}} extend one or more of these core structs.

Likewise, the required_capabilities GroupContext extension (defined in {{Section 11.1 of !RFC9420}} and reproduced below) contains all mandatory to support non-default extensions in its extension_types vector. Its proposal_types vector contains any mandatory to support Proposals. Its credential_types vector contains any mandatory credential types.

struct {
   ExtensionType extension_types<V>;
   ProposalType proposal_types<V>;
   CredentialType credential_types<V>;
} RequiredCapabilities;

Due to an oversight in {{!RFC9420}}, the Capabilities struct does not include MLS Wire Formats. Instead, this document defines two extensions: supported_wire_formats (which can appear in LeafNodes), and required_wire_formats (which can appear in the GroupContext).

struct {
   WireFormat wire_formats<V>;
} WireFormats

WireFormats supported_wire_formats;
WireFormats requires_wire_formats;

This document also defines new components of the app_data_dictionary extension for supported and required Safe AAD, media types, and components.

The safe_aad component contains a list of components IDs. When present (in an app_data_dictionary extension) in a LeafNode, the semantic is the list of supported components that use Safe AAD. When present (in an app_data_dictionary extension) in the GroupContext, the semantic is the list of required Safe AAD components (those that must be understood by the entire group). If the safe_aad component is present, even with an empty list, (in the app_data_dictionary extension) in the GroupContext, then the authenticated_data field always starts with the SafeAAD struct defined in {{safe-aad}}.

struct {
    ComponentID component_ids<V>;
} ComponentsList;

ComponentsList safe_aad;

The list of required and supported components follows the same model with the new component app_components. When present in a LeafNode, the semantic is the list of supported components. When present in the GroupContext, the semantic is the list of required components.

ComponentsList app_components;

Finally, the supported and required media types (formerly called MIME types) are communicated in the content_media_types component (see {{content-advertisement}}).

Extensions

AppAck

An AppAck object is used to acknowledge receipt of application messages. Though this information implies no change to the group, it is conveyed inside an AppEphermeral Proposal with a component ID app_ack, so that it is included in the group's transcript by being included in Commit messages.

struct {
    uint32 sender;
    uint32 first_generation;
    uint32 last_generation;
} MessageRange;

struct {
    MessageRange received_ranges<V>;
} AppAck;

An AppAck represents a set of messages received by the sender in the current epoch. Messages are represented by the sender and generation values in the MLSCiphertext for the message. Each MessageRange represents receipt of a span of messages whose generation values form a continuous range from first_generation to last_generation, inclusive.

AppAck objects are sent as a guard against the Delivery Service dropping application messages. The sequential nature of the generation field provides a degree of loss detection, since gaps in the generation sequence indicate dropped messages. AppAck completes this story by addressing the scenario where the Delivery Service drops all messages after a certain point, so that a later generation is never observed. Obviously, there is a risk that AppAck messages could be suppressed as well, but their inclusion in the transcript means that if they are suppressed then the group cannot advance at all.

Targeted messages

Description

MLS application messages make sending encrypted messages to all group members easy and efficient. Sometimes application protocols mandate that messages are only sent to specific group members, either for privacy or for efficiency reasons.

Targeted messages are a way to achieve this without having to create a new group with the sender and the specific recipients – which might not be possible or desired. Instead, targeted messages define the format and encryption of a message that is sent from a member of an existing group to another member of that group.

The goal is to provide a one-shot messaging mechanism that provides confidentiality and authentication, reusing mechanisms from {{!RFC9420}}, in particular {{!RFC9180}}.

Format

This extension uses the mls_extension_message WireFormat, where the content is a TargetedMessage.

struct {
  opaque group_id<V>;
  uint64 epoch;
  uint32 recipient_leaf_index;
  opaque authenticated_data<V>;
  opaque encrypted_sender_auth_data<V>;
  opaque hpke_ciphertext<V>;
} TargetedMessage;

enum {
  hpke_auth_psk(0),
  signature_hpke_psk(1),
} TargetedMessageAuthScheme;

struct {
  uint32 sender_leaf_index;
  TargetedMessageAuthScheme authentication_scheme;
  select (authentication_scheme) {
    case HPKEAuthPsk:
    case SignatureHPKEPsk:
      opaque signature<V>;
  }
  opaque kem_output<V>;
} TargetedMessageSenderAuthData;

struct {
  opaque group_id<V>;
  uint64 epoch;
  uint32 recipient_leaf_index;
  opaque authenticated_data<V>;
  TargetedMessageSenderAuthData sender_auth_data;
} TargetedMessageTBM;

struct {
  opaque group_id<V>;
  uint64 epoch;
  uint32 recipient_leaf_index;
  opaque authenticated_data<V>;
  uint32 sender_leaf_index;
  TargetedMessageAuthScheme authentication_scheme;
  opaque kem_output<V>;
  opaque hpke_ciphertext<V>;
} TargetedMessageTBS;

struct {
  opaque group_id<V>;
  uint64 epoch;
  opaque label<V> = "MLS 1.0 targeted message psk";
} PSKId;

Note that TargetedMessageTBS is only used with the TargetedMessageAuthScheme.SignatureHPKEPsk authentication mode.

Encryption

Targeted messages uses HPKE to encrypt the message content between two leaves.

Sender data encryption

In addition, TargetedMessageSenderAuthData is encrypted in a similar way to MLSSenderData as described in {{Section 6.3.2 of !RFC9420}}. The TargetedMessageSenderAuthData.sender_leaf_index field is the leaf index of the sender. The TargetedMessageSenderAuthData.authentication_scheme field is the authentication scheme used to authenticate the sender. The TargetedMessageSenderAuthData.signature field is the signature of the TargetedMessageTBS structure. The TargetedMessageSenderAuthData.kem_output field is the KEM output of the HPKE encryption.

The key and nonce provided to the AEAD are computed as the KDF of the first KDF.Nh bytes of the hpke_ciphertext generated in the following section. If the length of the hpke_ciphertext is less than KDF.Nh, the whole hpke_ciphertext is used. In pseudocode, the key and nonce are derived as:

sender_auth_data_secret
  = DeriveExtensionSecret(extension_secret, "targeted message sender auth data")

ciphertext_sample = hpke_ciphertext[0..KDF.Nh-1]

sender_data_key = ExpandWithLabel(sender_auth_data_secret, "key",
                      ciphertext_sample, AEAD.Nk)
sender_data_nonce = ExpandWithLabel(sender_auth_data_secret, "nonce",
                      ciphertext_sample, AEAD.Nn)

The Additional Authenticated Data (AAD) for the SenderAuthData ciphertext is the first three fields of TargetedMessage:

struct {
  opaque group_id<V>;
  uint64 epoch;
  uint32 recipient_leaf_index;
} SenderAuthDataAAD;

Padding

The TargetedMessage structure does not include a padding field. It is the responsibility of the sender to add padding to the message as used in the next section.

Authentication

For ciphersuites that support it, HPKE mode_auth_psk is used for authentication. For other ciphersuites, HPKE mode_psk is used along with a signature. The authentication scheme is indicated by the authentication_scheme field in TargetedMessageContent. See {{guidance-on-authentication-schemes}} for more information.

For the PSK part of the authentication, clients export a dedicated secret:

targeted_message_psk
  = DeriveExtensionSecret(extension_secret, "targeted message psk")

The functions SealAuth and OpenAuth defined in {{!RFC9180}} are used as described in {{safe-hpke}} with an empty context. Other functions are defined in {{!RFC9420}}.

Authentication with HPKE

The sender MUST set the authentication scheme to TargetedMessageAuthScheme.HPKEAuthPsk.

As described in {{safe-hpke}} the hpke_context is a LabeledExtensionContent struct with the following content, where group_context is the serialized context of the group.

label = "MLS 1.0 ExtensionData"
extension_type = ExtensionType
extension_data = group_context

The sender then computes the following:

(kem_output, hpke_ciphertext) = SealAuthPSK(receiver_node_public_key,
                                            hpke_context,
                                            targeted_message_tbm,
                                            message,
                                            targeted_message_psk,
                                            psk_id,
                                            sender_node_private_key)

The recipient computes the following:

message = OpenAuthPSK(kem_output,
                      receiver_node_private_key,
                      hpke_context,
                      targeted_message_tbm,
                      hpke_ciphertext,
                      targeted_message_psk,
                      psk_id,
                      sender_node_public_key)

Authentication with signatures

The sender MUST set the authentication scheme to TargetedMessageAuthScheme.SignatureHPKEPsk. The signature is done using the signature_key of the sender's LeafNode and the corresponding signature scheme used in the group.

The sender then computes the following with hpke_context defined as in {{authentication-with-hpke}}:

(kem_output, hpke_ciphertext) = SealPSK(receiver_node_public_key,
                                        hpke_context,
                                        targeted_message_tbm,
                                        message,
                                        targeted_message_psk,
                                        epoch)

The signature is computed as follows, where the extension_type is the type of this extension (see {{iana-considerations}}).

signature = SafeSignWithLabel(extension_type, ., "TargetedMessageTBS", targeted_message_tbs)

The recipient computes the following:

message = OpenPSK(kem_output,
                  receiver_node_private_key,
                  hpke_context,
                  targeted_message_tbm,
                  hpke_ciphertext,
                  targeted_message_psk,
                  epoch)

The recipient MUST verify the message authentication:

SafeVerifyWithLabel.verify(extension_type,
                        sender_leaf_node.signature_key,
                        "TargetedMessageTBS",
                        targeted_message_tbs,
                        signature)

Guidance on authentication schemes

If the group’s ciphersuite does not support HPKE mode_auth_psk, implementations MUST choose TargetedMessageAuthScheme.SignatureHPKEPsk.

If the group’s ciphersuite does support HPKE mode_auth_psk, implementations CAN choose TargetedMessageAuthScheme.HPKEAuthPsk if better efficiency and/or repudiability is desired. Implementations SHOULD consult {{Section 9.1.1 of !RFC9180}} beforehand.

Content Advertisement

Description

This section defines a minimal framing format so MLS clients can signal which media type is being sent inside the MLS application_data object when multiple formats are permitted in the same group.

It also defines a new content_media_types application component which is used to indicate support for specific formats, using the extensive IANA Media Types registry (formerly called MIME Types). When the content_media_types component is present (in the app_data_dictionary extension) in a LeafNode, it indicates that node's support for a particular (non-empty) list of media types. When the content_media_types component is present (in the app_data_dictionary extension) in the GroupContext, it indicates a (non-empty) list of media types that need to be supported by all members of that MLS group, and that the application_data will be framed using the application framing format described later in {{app-framing}}. This allows clients to confirm that all members of a group can communicate.

Note that when the membership of a group changes, or when the policy of the group changes, it is responsibility of the committer to insure that the membership and policies are compatible.

As clients are upgraded to support new formats they can use these extensions to detect when all members support a new or more efficient encoding, or select the relevant format or formats to send.

Vendor-specific media subtypes starting with vnd. can be registered with IANA without standards action as described in {{?RFC6838}}. Implementations which wish to send multiple formats in a single application message, may be interested in the multipart/alternative media type defined in {{?RFC2046}} or may use or define another type with similar semantics (for example using TLS Presentation Language syntax {{!RFC8446}}).

Note that the usage of IANA media types in general does not imply the usage of MIME Headers {{?RFC2045}} for framing.

Syntax

MediaType is a TLS encoding of a single IANA media type (including top-level type and subtype) and any of its parameters. Even if the parameter_value would have required formatting as a quoted-string in a text encoding, only the contents inside the quoted-string are included in parameter_value. Likewise, only the second character of a quoted-pair is included in parameter_value; the first escaping backslash ("") is omitted. MediaTypeList is an ordered list of MediaType objects.

struct {
    opaque parameter_name<V>;
    /* Note: parameter_value never includes the quotation marks of */
    /* an RFC 2045 quoted-string or the first "\" of a quoted-pair */
    opaque parameter_value<V>;
} Parameter;

struct {
    /* media_type is an IANA top-level media type, a "/" character,
     * and the IANA media subtype */
    opaque media_type<V>;

    /* a list of zero or more parameters defined for the subtype */
    Parameter parameters<V>;
} MediaType;

struct {
    /* must contain at least one item */
    MediaType media_types<V>;
} MediaTypeList;

MediaTypeList content_media_types;

Example IANA media types with optional parameters:

  image/png
  text/plain ;charset="UTF-8"
  application/json
  application/vnd.example.msgbus+cbor

For the example media type for text/plain, the media_type field would be text/plain, parameters would contain a single Parameter with a parameter_name of charset and a parameter_value of UTF-8.

Expected Behavior

An MLS client which implements this section SHOULD include the content_media_types component (in the app_data_dictionary extension) in its LeafNodes, listing all the media types it can receive. As usual, the client also includes content_media_types in the app_components list (in the app_data_dictionary extension) and support for the app_data_dictionary extension in its capabilities.extensions field in its LeafNodes (including in LeafNodes inside its KeyPackages).

When creating a new MLS group for an application using this specification, the group MAY include a content_media_types component (in the app_data_dictionary extension) in the GroupContext. (The creating client also includes its content_media_types component in its own LeafNode as described in the previous paragraph.)

MLS clients SHOULD NOT add an MLS client to an MLS group with content_media_types in its GroupContext unless the MLS client advertises it can support all of the required MediaTypes. As an exception, a client could be preconfigured to know that certain clients support the required types. Likewise, an MLS client is already forbidden from issuing or committing a GroupContextExtensions Proposal which introduces required extensions which are not supported by all members in the resulting epoch.

Framing of application_data {#app-framing}

When an MLS group contains the content_media_types component (in the app_data_dictionary extension) in its GroupContext, the application_data sent in that group is interpreted as ApplicationFraming as defined below:

  struct {
      MediaType media_type;
      opaque<V> application_content;
  } ApplicationFraming;

The media_type MAY be zero length, in which case, the media type of the application_content is interpreted as the first MediaType specified in the content_media_types component in the GroupContext.

SelfRemove Proposal

The design of the MLS protocol prevents a member of an MLS group from removing itself immediately from the group. (To cause an immediate change in the group, a member must send a Commit message. However the sender of a Commit message knows the keying material of the new epoch and therefore needs to be part of the group.) Instead a member wishing to remove itself can send a Remove Proposal and wait for another member to Commit its Proposal.

Unfortunately, MLS clients that join via an External Commit ignore pending, but otherwise valid, Remove Proposals. The member trying to remove itself has to monitor the group and send a new Remove Proposal in every new epoch until the member is removed. In a group with a burst of external joiners, a member connected over a high-latency link (or one that is merely unlucky) might have to wait several epochs to remove itself. A real-world situation in which this happens is a member trying to remove itself from a conference call as several dozen new participants are trying to join (often on the hour).

This section describes a new SelfRemove Proposal extension type. It is designed to be included in External Commits.

Extension Description

This document specifies a new MLS Proposal type called SelfRemove. Its syntax is described using the TLS Presentation Language {{!RFC8446}} below (its content is an empty struct). It is allowed in External Commits and requires an UpdatePath. SelfRemove proposals are only allowed in a Commit by reference. SelfRemove cannot be sent as an external proposal.

struct {} SelfRemove;

struct {
    ProposalType msg_type;
    select (Proposal.msg_type) {
        case add:                      Add;
        case update:                   Update;
        case remove:                   Remove;
        case psk:                      PreSharedKey;
        case reinit:                   ReInit;
        case external_init:            ExternalInit;
        case group_context_extensions: GroupContextExtensions;
        case self_remove:              SelfRemove;
    };
} Proposal;

The description of behavior below only applies if all the members of a group support this extension in their capabilities; such a group is a "self-remove-capable group".

An MLS client which supports this extension can send a SelfRemove Proposal whenever it would like to remove itself from a self-remove-capable group. Because the point of a SelfRemove Proposal is to be available to external joiners (which are not yet members), these proposals MUST be sent in an MLS PublicMessage.

Whenever a member receives a SelfRemove Proposal, it includes it along with any other pending Propsals when sending a Commit. It already MUST send a Commit of pending Proposals before sending new application messages.

When a member receives a Commit referencing one or more SelfRemove Proposals, it treats the proposal like a Remove Proposal, except the leaf node to remove is determined by looking in the Sender leaf_index of the original Proposal. The member is able to verify that the Sender was a member.

Whenever a new joiner is about to join a self-remove-capable group with an External Commit, the new joiner MUST fetch any pending SelfRemove Proposals along with the GroupInfo object, and include the SelfRemove Proposals in its External Commit by reference. (An ExternalCommit can contain zero or more SelfRemove proposals). The new joiner MUST validate the SelfRemove Proposal before including it by reference, except that it skips the validation of the membership_tag because a non-member cannot verify membership.

During validation, SelfRemove proposals are processed after Update proposals and before Remove proposals. If there is a pending SelfRemove proposal for a specific leaf node and a pending Remove proposal for the same leaf node, the Remove proposal is invalid. A client MUST NOT issue more than one SelfRemove proposal per epoch.

The MLS Delivery Service (DS) needs to validate SelfRemove Proposals it receives (except that it cannot validate the membership_tag). If the DS provides a GroupInfo object to an external joiner, the DS SHOULD attach any SelfRemove proposals known to the DS to the GroupInfo object.

As with Remove proposals, clients need to be able to receive a Commit message which removes them from the group via a SelfRemove. If the DS does not forward a Commit to a removed client, it needs to inform the removed client out-of-band.

Last resort KeyPackages

Type: KeyPackage extension

Description

{{Section 10 of !RFC9420}} details that clients are required to pre-publish KeyPackages so that other clients can add them to groups asynchronously. It also states that they should not be re-used:

KeyPackages are intended to be used only once and SHOULD NOT be reused except in the case of a "last resort" KeyPackage (see Section 16.8). Clients MAY generate and publish multiple KeyPackages to support multiple cipher suites.

{{Section 16.8 of !RFC9420}} then introduces the notion of last-resort KeyPackages as follows:

An application MAY allow for reuse of a "last resort" KeyPackage in order to prevent denial-of-service attacks.

However, {{!RFC9420}} does not specify how to distinguish regular KeyPackages from last-resort ones. The last_resort_key_package KeyPackage application component defined in this section fills this gap and allows clients to specifically mark KeyPackages as KeyPackages of last resort that MAY be used more than once in scenarios where all other KeyPackages have already been used.

The component allows clients that pre-publish KeyPackages to signal to the Delivery Service which KeyPackage(s) are meant to be used as last resort KeyPackages.

An additional benefit of using a component rather than communicating the information out-of-band is that the component is still present in Add proposals. Clients processing such Add proposals can authenticate that a KeyPackage is a last-resort KeyPackage and MAY make policy decisions based on that information.

Format

The purpose of the application component is simply to mark a given KeyPackage, which means it carries no additional data.

As a result, a LastResort Extension contains the component_id with an empty data field.

Multi-Credentials

Multi-credentials address use cases where there might not be a single credential that captures all of a client's authenticated attributes. For example, an enterprise messaging client may wish to provide attributes both from its messaging service, to prove that its user has a given handle in that service, and from its corporate owner, to prove that its user is an employee of the corporation. Multi-credentials can also be used in migration scenarios, where some clients in a group might wish to rely on a newer type of credential, but other clients haven't yet been upgraded.

New credential types MultiCredential and WeakMultiCredential are defined as shown below. These credential types are indicated with the values multi and weak-multi (see {{iana-creds}}).

struct {
  CipherSuite cipher_suite;
  Credential credential;
  SignaturePublicKey credential_key;

  /* SignWithLabel(., "CredentialBindingTBS", CredentialBindingTBS) */
  opaque signature<V>;
} CredentialBinding

struct {
  CredentialBinding bindings<V>;
} MultiCredential;

struct {
  CredentialBinding bindings<V>;
} WeakMultiCredential;

The two types of credentials are processed in exactly the same way. The only difference is in how they are treated when evaluating support by other clients, as discussed below.

Credential Bindings

A multi-credential consists of a collection of "credential bindings". Each credential binding is a signed statement by the holder of the credential that the signature key in the LeafNode belongs to the holder of that credential. Specifically, the signature is computed using the MLS SignWithLabel function, with label "CredentialBindingTBS" and with a content that covers the contents of the CredentialBinding, plus the signature_key field from the LeafNode in which this credential will be embedded.

struct {
  CipherSuite cipher_suite;
  Credential credential;
  SignaturePublicKey credential_key;
  SignaturePublicKey signature_key;
} CredentialBindingTBS;

The cipher_suite for a credential is NOT REQUIRED to match the cipher suite for the MLS group in which it is used, but MUST meet the support requirements with regard to support by group members discussed below.

Verifying a Multi-Credential

A credential binding is supported by a client if the client supports the credential type and cipher suite of the binding. A credential binding is valid in the context of a given LeafNode if both of the following are true:

  • The credential is valid according to the MLS Authentication Service.

  • The credential_key corresponds to the specified credential, in the same way that the signature_key would have to correspond to the credential if the credential were presented in a LeafNode.

  • The signature field is valid with respect to the signature_key value in the leaf node.

A client that receives a credential of type multi in a LeafNode MUST verify that all of the following are true:

  • All members of the group support credential type multi.

  • For each credential binding in the multi-credential:

    • Every member of the group supports the cipher suite and credential type values for the binding.

    • The binding is valid in the context of the LeafNode.

A client that receives a credential of type weak-multi in a LeafNode MUST verify that all of the following are true:

  • All members of the group support credential type weak-multi.

  • Each member of the group supports at least one binding in the multi-credential. (Different members may support different subsets.)

  • Every binding that this client supports is valid in the context of the LeafNode.

IANA Considerations

This document requests the addition of various new values under the heading of "Messaging Layer Security". Each registration is organized under the relevant registry Type.

This document also requests the creation of a new MLS applications components registry as described in {{iana-components}}.

RFC EDITOR: Please replace XXXX throughout with the RFC number assigned to this document

MLS Wire Formats

MLS Targeted Message

The mls_targeted_message MLS Wire Format is used to send a message to a subset of members of an MLS group.

  • Value: 0x0006 (suggested)
  • Name: mls_targeted_message
  • Recommended: Y
  • Reference: RFC XXXX

MLS Extension Types

app_data_dictionary MLS Extension

The app_data_dictionary MLS Extension Type is used inside KeyPackage, LeafNode, GroupContext, or GroupInfo objects. It contains a sorted list of application component data objects (at most one per component).

  • Value: 0x0006 (suggested)
  • Name: app_data_dictionary
  • Message(s): KP: This extension may appear in KeyPackage objects LN: This extension may appear in LeafNode objects GC: This extension may appear in GroupContext objects GI: This extension may appear in GroupInfo objects
  • Recommended: Y
  • Reference: RFC XXXX

supported_wire_formats MLS Extension

The supported_wire_formats MLS Extension Type is used inside LeafNode objects. It contains a list of non-default Wire Formats supported by the client node.

  • Value: 0x0007 (suggested)
  • Name: supported_wire_formats
  • Message(s): LN: This extension may appear in LeafNode objects
  • Recommended: Y
  • Reference: RFC XXXX

required_wire_formats MLS Extension

The required_wire_formats MLS Extension Type is used inside GroupContext objects. It contains a list of non-default Wire Formats that are mandatory for all MLS members of the group to support.

  • Value: 0x0008 (suggested)
  • Name: required_wire_formats
  • Message(s): GC: This extension may appear in GroupContext objects
  • Recommended: Y
  • Reference: RFC XXXX

targeted_messages_capability MLS Extension

The targeted_messages_capability MLS Extension Type is used in the capabilities.extensions field of LeafNodes to indicate the support for the Targeted Messages Extension, and in the required_capabilities.extension_types field of the GroupContext to indicate all members of the group must support it. The extension does not carry any payload.

  • Value: 0x0009 (suggested)
  • Name: targeted_messages_capability
  • Message(s): LN: This extension may appear in LeafNode objects GC: This extension may appear in GroupContext objects
  • Recommended: Y
  • Reference: RFC XXXX

MLS Proposal Types

AppDataUpdate Proposal

The app_data_update MLS Proposal Type is used to efficiently update application component data stored in the app_data_dictionary GroupContext extension.

  • Value: 0x0008 (suggested)
  • Name: app_data_update
  • Recommended: Y
  • External: Y
  • Path Required: N

AppEphemeral Proposal

The app_ephemeral MLS Proposal Type is used to send opaque ephemeral application data that needs to be synchronized with a specific MLS epoch.

  • Value: 0x0009 (suggested)
  • Name: app_ephemeral
  • Recommended: Y
  • External: Y
  • Path Required: N

SelfRemove Proposal

The self_remove MLS Proposal Type is used for a member to remove itself from a group more efficiently than using a remove proposal type, as the self_remove type is permitted in External Commits.

  • Value: 0x0008 (suggested)
  • Name: self_remove
  • Recommended: Y
  • External: N
  • Path Required: Y

MLS Credential Types {#iana-creds}

Multi Credential

  • Value: 0x0003 (suggested)
  • Name: multi
  • Recommended: Y
  • Reference: RFC XXXX

Weak Multi Credential

  • Value: 0x0004
  • Name: weak-multi
  • Recommended: Y
  • Reference: RFC XXXX

MLS Component Types {#iana-components}

This document requests the creation of a new IANA "MLS Component Types" registry under the "Messaging Layer Security" group registry heading. Assignments to this registry in the range 0x0000 0000 to 0x7FFF FFFF are via Specification Required policy {{!RFC8126}} using the MLS Designated Experts. Assignments in the range 0x8000 0000 to 0xFFFF FFFF are for private use.

Template:

  • Value: The numeric value of the component ID
  • Name: The name of the component
  • Where: The objects(s) in which the component may appear, drawn from the following list:
    • AD: SafeAAD objects
    • AE: AppEpheral proposals
    • ES: Exporter Secret labels
    • GC: GroupContext objects
    • GI: GroupInfo objects
    • HP: HPKE key labels
    • KP: KeyPackage objects
    • LN: LeafNode objects
    • PS: PSK labels
    • SK: Signature Key labels
  • Recommended: Same as in {{Section 17.1 of !RFC9420}}
  • Reference: The document where this component is defined

The restrictions noted in the "Where" column are to be enforced by the application. MLS implementations MUST NOT impose restrictions on where component IDs are used in which parts of MLS, unless specifically directed to by the application.

Initial Contents:

| Value | Name | Where | R | Ref | |---------------+--------------------------+-------+---+---------| | 0x0000 0000 | RESERVED | N/A | - | RFCXXXX | | 0x0000 0001 | app_components | LN,GC | Y | RFCXXXX | | 0x0000 0002 | safe_aad | LN,GC | Y | RFCXXXX | | 0x0000 0003 | content_media_types | LN,GC | Y | RFCXXXX | | 0x0000 0004 | last_resort_key_package | KP | Y | RFCXXXX | | 0x0000 0005 | app_ack | AE | Y | RFCXXXX | | 0x8000 0000 - 0xFFFF FFFF | Reserved for Private Use | N/A | N | RFCXXXX |

Security considerations

Safe Application API

The Safe Application API provides the following security guarantee: If an application uses MLS with application components, the security guarantees of the base MLS protocol and the security guarantees of each application component analyzed in isolation, still hold for the composed application of the MLS protocol. In other words, the Safe Application API protects applications from careless component developers. It is not possible that a combination of components (the developers of which did not know about each other) impedes the security of the base MLS protocol or any other component. No further analysis of the combination is necessary. This also means that any security vulnerabilities introduced by one component do not spread to other components or the base MLS implementation.

AppAck

When AppAck objects are received, they allow clients to detect if the Delivery Service (or an intermediary) dropped application messages, since gaps in the generation sequence indicate dropped messages. When AppAck messages are accepted by the Delivery Service, but not received by some members, the members who have missed the corresponding AppEphemeral proposals will not be able to send or receive a commit message, because the proposal is included in the transcript hash. Likewise if AppAck objects and/or commits are sent periodically by every member, other members will be able to detect a member that is no longer sending on that schedule or whose handshake messages are being suppressed by the DS.

Note: External Commits do not typically contain pending proposals (including AppEphemeral proposals). Client that send an AppAck component in an AppEphemeral proposal will need to send a new AppAck component in an AppEphemeral proposal (in the new epoch) after receiving an External Commit until it has been incorporated into an accepted Commit.

The schedule on which AppAck objects are sent in AppEphemeral proposals is up to the application,and determines which cases of loss/suppression are detected. For example:

  • The application might have the committer include an AppAck whenever a Commit is sent, so that other members could know when one of their messages did not reach the committer.

  • The application could have a client send an AppAck whenever an application message is sent, covering all messages received since its last AppAck. This would provide a complete view of any losses experienced by active members.

  • The application could simply have clients send AppAck proposals on a timer, so that all participants' state would be known.

An application using AppAck to guard against loss/suppression of application messages also needs to ensure that AppAck messages and the Commits that reference them are not dropped. One way to do this is to always encrypt Proposal and Commit messages, to make it more difficult for the Delivery Service to recognize which messages contain AppAcks. The application can also have clients enforce an AppAck schedule, reporting loss if an AppAck is not received at the expected time.

Targeted Messages

In addition to the sender authentication, Targeted Messages are authenticated by using a preshared key (PSK) between the sender and the recipient. The PSK is exported from the group key schedule using the label "targeted message psk". This ensures that the PSK is only valid for a specific group and epoch, and the Forward Secrecy and Post-Compromise Security guarantees of the group key schedule apply to the targeted messages as well. The PSK also ensures that an attacker needs access to the private group state in addition to the HPKE/signature's private keys. This improves confidentiality guarantees against passive attackers and authentication guarantees against active attackers.

Content Advertisement

Use of the content_media_types component could leak some private information visible in KeyPackages and inside an MLS group. This could be used to infer a specific implementation, platform, or even version. Clients should carefully consider the privacy implications in their environment of making a list of acceptable media types available.

Implementations need to be prepared to parse media types containing long parameter lists, potentially containing characters which would be escaped or quoted in {{!RFC5322}}.

SelfRemove

An external recipient of a SelfRemove Proposal cannot verify the membership_tag. However, an external joiner also has no way to completely validate a GroupInfo object that it receives. An insider can prevent an External Join by providing either an invalid GroupInfo object or an invalid SelfRemove Proposal. The security properties of external joins does not change with the addition of this proposal type.

Multi Credentials

Using a Weak Multi Credential reduces the overall credential security to the security of the least secure of its credential bindings.

--- back

Change Log

RFC EDITOR PLEASE DELETE THIS SECTION.

draft-06

  • Integrate notion of Application API from draft-barnes-mls-appsync

draft-05

  • Include definition of ExtensionState extension
  • Add safe use of AAD to Safe Extensions framework
  • Clarify how capabilities negotiation works in Safe Extensions framework

draft-04

  • No changes (prevent expiration)

draft-03

  • Add Last Resort KeyPackage extension
  • Add Safe Extensions framework
  • Add SelfRemove Proposal

draft-02

  • No changes (prevent expiration)

draft-01

  • Add Content Advertisement extensions

draft-00

  • Initial adoption of draft-robert-mls-protocol-00 as a WG item.
  • Add Targeted Messages extension (*)

Old Safe Extensions Text

The MLS specification is extensible in a variety of ways (see {{Section 13 of !RFC9420}}) and describes the negotiation and other handling of extensions and their data within the protocol. However, it does not provide guidance on how extensions can or should safely interact with the base MLS protocol. The goal of this section is to simplify the task of developing MLS extensions.

Extension state: anchoring, storage and agreement

The safe extension framework can help an MLS extension ensure that all group members agree on a piece of extension-specific state by using the ExtensionState GroupContext extension. The ownership of an ExtensionState extension in the context of the safe extension framework is determined by the extension_type field. The extension with a matching extension_type is called the owning extension.

enum {
  reserved(0),
  read(1),
  none(2),
 (255)
} Permissions;

enum {
  reserved(0),
  hash(1),
  data(2),
} HashOrData;

struct {
  HashOrData hash_or_data;
  select(hash_or_data) {
    case hash:
      HashReference state_hash;
    case data:
      opaque state<V>;
  }
} ExtensionPayload;

struct {
  extensionType extension_type;
  Permissions read;
  ExtensionPayload payload;
} ExtensionState;

The ExtensionState GroupContext extension contains data either directly (if hash_or_data = data) or inditectly via a hash (if hash_or_data = hash).

The owning extension can read and write the state stored in an ExtensionState extension using an extension-defined proposal (see ). The semantics of the proposal determines how the state is changed.

The read variable determines the permissions that other MLS extensions have w.r.t. the data stored within. read allows other MLS extensions to read that data via their own proposals, while none marks the data as private to the owning MLS extension.

Other extensions may never write to the ExtensionState of the owning MLS extension.

Direct vs. hash-based storage

Storing the data directly in the ExtensionState means the data becomes part of the group state. Depending on the application design, this can be advantageous, because it is distributed via Welcome messages. However, it could also mean that the data is visible to the delivery service. Additionally, if the application makes use of GroupContextExtension proposals, it may be necessary to send all of the data with each such extension.

Including the data by hash only allows group members to agree on the data indirectly, relying on the collision resistance of the associated hash function. The data itself, however, may have to be transmitted out-of-band to new joiners.

Extension Design Guidance

While extensions can modify the protocol flow of MLS and the associated properties in arbitrary ways, the base MLS protocol already enables a number of functionalities that extensions can use without modifying MLS itself. Extension authors should consider using these built-in mechanisms before employing more intrusive changes to the protocol.