The Messaging Layer Security (MLS) ProtocolCiscorlb@ipv.sxFacebookjmillican@fb.comGoogleemadomara@google.comUniversity of Oxfordme@katriel.co.ukWireraphael@wire.com
Security
Internet-DraftMessaging applications are increasingly making use of end-to-end
security mechanisms to ensure that messages are only accessible to
the communicating endpoints, and not to any servers involved in delivering
messages. Establishing keys to provide such protections is
challenging for group chat settings, in which more than two
clients need to agree on a key but may not be online at the same
time. In this document, we specify a key establishment
protocol that provides efficient asynchronous group key establishment
with forward secrecy and post-compromise security for groups
in size ranging from two to thousands.DISCLAIMER: This is a work-in-progress draft of MLS and has not yet
seen significant security analysis. It should not be used as a basis
for building production systems.RFC EDITOR: PLEASE REMOVE THE FOLLOWING PARAGRAPH The source for
this draft is maintained in GitHub. Suggested changes should be
submitted as pull requests at https://github.com/mlswg/mls-protocol.
Instructions are on that page as well. Editorial changes can be
managed in GitHub, but any substantive change should be discussed on
the MLS mailing list.A group of users who want to send each other encrypted messages needs
a way to derive shared symmetric encryption keys. For two parties,
this problem has been studied thoroughly, with the Double Ratchet
emerging as a common solution .
Channels implementing the Double Ratchet enjoy fine-grained forward secrecy
as well as post-compromise security, but are nonetheless efficient
enough for heavy use over low-bandwidth networks.For a group of size greater than two, a common strategy is to
unilaterally broadcast symmetric “sender” keys over existing shared
symmetric channels, and then for each member to send messages to the
group encrypted with their own sender key. Unfortunately, while this
improves efficiency over pairwise broadcast of individual messages and
provides forward secrecy (with the addition of a hash ratchet),
it is difficult to achieve post-compromise security with
sender keys. An adversary who learns a sender key can often indefinitely and
passively eavesdrop on that member’s messages. Generating and
distributing a new sender key provides a form of post-compromise
security with regard to that sender. However, it requires
computation and communications resources that scale linearly with
the size of the group.In this document, we describe a protocol based on tree structures
that enable asynchronous group keying with forward secrecy and
post-compromise security. Based on earlier work on “asynchronous
ratcheting trees” , the mechanism presented here use a
asynchronous key-encapsulation mechanism for tree structures.
This mechanism allows the members of the group to derive and update
shared keys with costs that scale as the log of the group size.RFC EDITOR PLEASE DELETE THIS SECTION.draft-04Updating the language to be similar to the Architecture documentECIES is now renamed in favor of HPKE (*)Using a KDF instead of a Hash in TreeKEM (*)draft-03Added ciphersuites and signature schemes (*)Re-ordered fields in UserInitKey to make parsing easier (*)Fixed inconsistencies between Welcome and GroupState (*)Added encryption of the Welcome message (*)draft-02Removed ART (*)Allowed partial trees to avoid double-joins (*)Added explicit key confirmation (*)draft-01Initial description of the Message Protection mechanism. (*)Initial specification proposal for the Application Key Schedule
using the per-participant chaining of the Application Secret design. (*)Initial specification proposal for an encryption mechanism to protect
Application Messages using an AEAD scheme. (*)Initial specification proposal for an authentication mechanism
of Application Messages using signatures. (*)Initial specification proposal for a padding mechanism to improving
protection of Application Messages against traffic analysis. (*)Inversion of the Group Init Add and Application Secret derivations
in the Handshake Key Schedule to be ease chaining in case we switch
design. (*)Removal of the UserAdd construct and split of GroupAdd into Add
and Welcome messages (*)Initial proposal for authenticating Handshake messages by signing
over group state and including group state in the key schedule (*)Added an appendix with example code for tree mathChanged the ECIES mechanism used by TreeKEM so that it uses nonces
generated from the shared secretdraft-00Initial adoption of draft-barnes-mls-protocol-01 as a WG item.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 when, and only when, they appear in all
capitals, as shown here.
An agent that uses this protocol to establish shared cryptographic
state with other clients. A client is defined by the
cryptographic keys it holds. An application or user may use one client
per device (keeping keys local to each device) or sync keys among
a user’s devices so that each user appears as a single client.
A collection of clients with shared cryptographic state.
A client that is included in the shared state of a group, hence
has access to the group’s secrets.
A short-lived HPKE key pair used to introduce a new
client to a group. Initialization keys are published for
each client (UserInitKey).
A secret that represent a member’s contribution to the group secret
(so called because the members’ leaf keys are the leaves in the
group’s ratchet tree).
A long-lived signing key pair used to authenticate the sender of a
message.Terminology specific to tree computations is described in
.We use the TLS presentation language to
describe the structure of protocol messages.This protocol is designed to execute in the context of a Messaging Service (MS)
as described in [I-D.ietf-mls-architecture]. In particular, we assume
the MS provides the following services:A long-term identity key provider which allows clients to authenticate
protocol messages in a group. These keys MUST be kept for the lifetime of the
group as there is no mechanism in the protocol for changing a client’s
identity key.A broadcast channel, for each group, which will relay a message to all members
of a group. For the most part, we assume that this channel delivers messages
in the same order to all participants. (See for further
considerations.)A directory to which clients can publish initialization keys and download
initialization keys for other participants.The goal of this protocol is to allow a group of clients to exchange confidential and
authenticated messages. It does so by deriving a sequence of secrets and keys known only to members. Those
should be secret against an active network adversary and should have both forward and
post-compromise secrecy with respect to compromise of a participant.We describe the information stored by each client as a state, which includes both public and
private data. An initial state, including an initial set of clients, is set up by a group
creator using the Init algorithm and based on information pre-published by clients. The creator
sends the Init message to the clients, who can then set up their own group state and derive
the same shared secret. Clients then exchange messages to produce new shared states which are
causally linked to their predecessors, forming a logical Directed Acyclic Graph (DAG) of states.
Members can send Update messages for post-compromise secrecy and new clients can be
added or existing members removed from the group.The protocol algorithms we specify here follow. Each algorithm specifies both (i) how a client
performs the operation and (ii) how other clients update their state based on it.There are four major operations in the lifecycle of a group:Adding a member, initiated by a current member;Updating the leaf secret of a member;Removing a member.Before the initialization of a group, clients publish UserInitKey
objects to a directory provided to the Messaging Service.When a client A wants to establish a group with B and C, it
first downloads UserInitKeys for B and C. It then initializes a group state
containing only itself and uses the UserInitKeys to compute Welcome and Add messages
to add B and C, in a sequence chosen by A. The Welcome messages are
sent directly to the new members (there is no need to send them to
the group).
The Add messages are broadcasted to the group, and processed in sequence
by B and C. Messages received before a client has joined the
group are ignored. Only after A has received its Add messages
back from the server does it update its state to reflect their addition.Subsequent additions of group members proceed in the same way. Any
member of the group can download an UserInitKey for a new client
and broadcast an Add message that the current group can use to update
their state and the new client can use to initialize its state.To enforce forward secrecy and post-compromise security of messages,
each member periodically updates its leaf secret which represents
its contribution to the group secret. Any member of the
group can send an Update at any time by generating a fresh leaf secret
and sending an Update message that describes how to update the
group secret with that new information. Once all members have
processed this message, the group’s secrets will be unknown to an
attacker that had compromised the sender’s prior leaf secret.It is left to the application to determine the interval of time between
Update messages. This policy could require a change for each message, or
it could require sending an update every week or more.Members are removed from the group in a similar way, as an update
is effectively removing the old leaf from the group.
Any member of the group can generate a Remove message that adds new
entropy to the group state that is known to all members except the
removed member. After other participants have processed this message,
the group’s secrets will be unknown to the removed participant.
Note that this does not necessarily imply that any member
is actually allowed to evict other members; groups can layer
authentication-based access control policies on top of these
basic mechanism.The protocol uses “ratchet trees” for deriving shared secrets among
a group of clients.Trees consist of nodes. A node is a
leaf if it has no children, and a parent otherwise; note that all
parents in our trees have precisely
two children, a left child and a right child. A node is the root
of a tree if it has no parents, and intermediate if it has both
children and parents. The descendants of a node are that node, its
children, and the descendants of its children, and we say a tree
contains a node if that node is a descendant of the root of the
tree. Nodes are siblings if they share the same parent.A subtree of a tree is the tree given by the descendants of any
node, the head of the subtree. The size of a tree or subtree is the
number of leaf nodes it contains. For a given parent node, its left
subtree is the subtree with its left child as head (respectively
right subtree).All trees used in this protocol are left-balanced binary trees. A
binary tree is full (and balanced) if its size is a power of
two and for any parent node in the tree, its left and right subtrees
have the same size. If a subtree is full and it is not a subset of
any other full subtree, then it is maximal.A binary tree is left-balanced if for every
parent, either the parent is balanced, or the left subtree of that
parent is the largest full subtree that could be constructed from
the leaves present in the parent’s own subtree. Note
that given a list of n items, there is a unique left-balanced
binary tree structure with these elements as leaves. In such a
left-balanced tree, the k-th leaf node refers to the k-th leaf
node in the tree when counting from the left, starting from 0.The direct path of a root is the empty list, and of any other node
is the concatenation of that node with the direct path of its
parent. The copath of a node is the list of siblings of nodes in its
direct path, excluding the root. The frontier of a tree is the list of heads of the maximal
full subtrees of the tree, ordered from left to right.For example, in the below tree:The direct path of C is (C, CD, ABCD)The copath of C is (D, AB, EFG)The frontier of the tree is (ABCD, EF, G)Each node in the tree is assigned an index, starting at zero and
running from left to right. A node is a leaf node if and only if it
has an even index. The indices for the nodes in the above tree are
as follows:0 = A1 = AB2 = B3 = ABCD4 = C5 = CD6 = D7 = ABCDEFG8 = E9 = EF10 = F11 = EFG12 = G(Note that left-balanced binary trees are the same structure that is
used for the Merkle trees in the Certificate Transparency protocol
.)Ratchet trees are used for generating shared group secrets. In this
section, we describe the structure of a ratchet tree. A particular
instance of a ratchet tree is based on the following cryptographic
primitives, defined by the ciphersuite in use:A Diffie-Hellman finite-field group or elliptic curveA Key Derivation Function (KDF)A Derive-Key-Pair function that produces an asymmetric keypair
from a node secretA ratchet tree is a left-balanced binary tree, in which each node
contains up to three values:A secret octet string (optional)An asymmetric private key (optional)An asymmetric public keyThe contents of the parent are based on the latest-updated child.
Nodes in a tree are always updated along the “direct path” from a
leaf to the root. The generator of the update chooses a random
secret value “path_secret[0]”, and generates a sequence of “path
secrets”, one for each node from the leaf to the root. That is,
path_secret[0] is used for the leaf, path_secret[1] for its parent,
and so on. At each step, the path secret is used to derive a new
secret value for the corresponding node, from which the node’s key
pair is derived.For example, suppose there is a group with four participants:If the first participant subsequently generates an update based on a
secret X, then the sender would generate the following sequence of
path secrets and node secrets:After the update, the tree will have the following structure, where
“ns[i]” represents the node_secret values generated as described
above:A node in the tree may be blank, indicating that no value is
present at that node. The resolution of a node is an ordered list
of non-blank nodes that collectively cover all non-blank descendants
of the node. The nodes in a resolution are ordered according to
their indices.The resolution of a non-blank node is a one element list
containing the node itselfThe resolution of a blank leaf node is the empty listThe resolution of a blank intermediate node is the result of
concatinating the resolution of its left child with the resolution
of its right child, in that orderFor example, consider the following tree, where the “_” character
represents a blank node:In this tree, we can see all three of the above rules in play:The resolution of node 5 is the list [CD]The resolution of node 2 is the empty list []The resolution of node 3 is the list [A, CD]In order to update the state of the group such as adding and
removing clients, MLS messages are used to make changes to the
group’s ratchet tree. The member proposing an update to the
tree transmits a set of values for intermediate nodes in the
direct path of a leaf. Other members in the group
can use these nodes to update their view of the tree, aligning their
copy of the tree to the sender’s.To perform an update for a leaf, the sender transmits the following
information for each node in the direct path of the leaf:The public key for the nodeZero or more encrypted copies of the node’s parent secret valueThe secret value is encrypted for the subtree corresponding to the
parent’s non-updated child, i.e., the child on the copath of the leaf node.
There is one encrypted secret for each public key in the resolution
of the non-updated child. In particular, for the leaf node, there
are no encrypted secrets, since a leaf node has no children.The recipient of an update processes it with the following steps:Compute the updated secret values
* Identify a node in the direct path for which the local member
is in the subtree of the non-updated child
* Identify a node in the resolution of the copath node for
which this node has a private key
* Decrypt the secret value for the parent of the copath node using
the private key from the resolution node
* Derive secret values for ancestors of that node using the KDF keyed with the
decrypted secretMerge the updated secrets into the tree
* Replace the public keys for nodes on the direct path with the
received public keys
* For nodes where an updated secret was computed in step 1,
replace the secret value for the node with the updated valueFor example, suppose we had the following tree:If an update is made along the direct path B-E-G, then the following
values will be transmitted (using pk(X) to represent the public key
corresponding to the secret value X and E(K, S) to represent
public-key encryption to the public key K of the secret value S):Public KeyCiphertext(s)pk(G)E(pk(C), G), E(pk(D), G)pk(E)E(pk(A), E)pk(B)Each MLS session uses a single ciphersuite that specifies the
following primitives to be used in group key computations:A hash functionA Diffie-Hellman finite-field group or elliptic curveAn AEAD encryption algorithm The ciphersuite must also specify an algorithm Derive-Key-Pair
that maps octet strings with the same length as the output of the
hash function to key pairs for the asymmetric encryption scheme.Public keys used in the protocol are opaque values
in a format defined by the ciphersuite, using the following types:Cryptographic algorithms are indicated using the following types:This ciphersuite uses the following primitives:Hash function: SHA-256Diffie-Hellman group: Curve25519 AEAD: AES-128-GCMGiven an octet string X, the private key produced by the
Derive-Key-Pair operation is SHA-256(X). (Recall that any 32-octet
string is a valid Curve25519 private key.) The corresponding public
key is X25519(SHA-256(X), 9).Implementations SHOULD use the approach
specified in to calculate the Diffie-Hellman shared secret.
Implementations MUST check whether the computed Diffie-Hellman shared
secret is the all-zero value and abort if so, as described in
Section 6 of . If implementers use an alternative
implementation of these elliptic curves, they SHOULD perform the
additional checks specified in Section 7 of This ciphersuite uses the following primitives:Hash function: SHA-256Diffie-Hellman group: secp256r1 (NIST P-256)AEAD: AES-128-GCMGiven an octet string X, the private key produced by the
Derive-Key-Pair operation is SHA-256(X), interpreted as a big-endian
integer. The corresponding public key is the result of multiplying
the standard P-256 base point by this integer.P-256 ECDH calculations (including parameter
and key generation as well as the shared secret calculation) are
performed according to using the ECKAS-DH1 scheme with the identity
map as key derivation function (KDF), so that the shared secret is the
x-coordinate of the ECDH shared secret elliptic curve point represented
as an octet string. Note that this octet string (Z in IEEE 1363 terminology)
as output by FE2OSP, the Field Element to Octet String Conversion
Primitive, has constant length for any given field; leading zeros
found in this octet string MUST NOT be truncated.(Note that this use of the identity KDF is a technicality. The
complete picture is that ECDH is employed with a non-trivial KDF
because MLS does not directly use this secret for anything
other than for computing other secrets.)Clients MUST validate remote public values by ensuring
that the point is a valid point on the elliptic curve.
The appropriate validation procedures are defined in Section 4.3.7 of
and alternatively in Section 5.6.2.3 of .
This process consists of three steps: (1) verify that the value 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.A member of a group authenticates the identities of other
participants by means of credentials issued by some authentication
system, e.g., a PKI. Each type of credential MUST express the
following data:The public key of a signature key pairThe identity of the holder of the private keyThe signature scheme that the holder will use to sign MLS messagesCredentials MAY also include information that allows a relying party
to verify the identity / signing key binding.Each member of the group maintains a representation of the
state of the group:The fields in this state have the following semantics:The group_id field is an application-defined identifier for the
group.The epoch field represents the current version of the group key.The roster field contains credentials for the occupied slots in
the tree, including the identity and signature public key for the
holder of the slot.The tree field contains the public keys corresponding to the
nodes of the ratchet tree for this group. The length of this
vector MUST be 2*size - 1, where size is the length of the
roster, since this is the number of nodes in a tree with size
leaves, according to the structure described in .The transcript_hash field contains the list of GroupOperation
messages that led to this state.When a new member is added to the group, an existing member of the
group provides the new member with a Welcome message. The Welcome
message provides the information the new member needs to initialize
its GroupState.Different group operations will have different effects on the group
state. These effects are described in their respective subsections
of . The following rules apply to all
operations:The group_id field is constantThe epoch field increments by one for each GroupOperation that
is processedThe transcript_hash is updated by a GroupOperation message
operation in the following way:When a new one-member group is created (which requires no
GroupOperation), the transcript_hash field is set to an all-zero
vector of length Hash.length, where the Hash algorithm is defined
by the ciphersuite.As described in , each MLS message needs to
transmit node values along the direct path of a leaf.
The path contains a public key for the leaf node, and a
public key and encrypted secret value for intermediate nodes in the
path. In both cases, the path is ordered from the leaf to the root;
each node MUST be the parent of its predecessor.The length of the node\_secrets vector MUST be zero for the first
node in the path. For the remaining elements in the vector, the
number of ciphertexts in the node\_secrets vector MUST be equal to
the length of the resolution of the corresponding copath node. Each
ciphertext in the list is the encryption to the corresponding node
in the resolution.The HPKECiphertext values are computed according to the Encrypt
function defined in .Decryption is performed in the corresponding way, using the private
key of the resolution node and the ephemeral public key
transmitted in the message.Group keys are derived using the HKDF-Extract and HKDF-Expand
functions as defined in , as well as the functions
defined below:The Hash function used by HKDF is the ciphersuite hash algorithm.
Hash.length is its output length in bytes. In the below diagram:HKDF-Extract takes its salt argument from the top and its IKM
argument from the leftDerive-Secret takes its Secret argument from the incoming arrowWhen processing a handshake message, a client combines the
following information to derive new epoch secrets:The init secret from the previous epochThe update secret for the current epochThe GroupState object for current epochGiven these inputs, the derivation of secrets for an epoch
proceeds as shown in the following diagram:In order to facilitate asynchronous addition of clients to a
group, it is possible to pre-publish initialization keys that
provide some public information about a user. UserInitKey
messages provide information about a client that any existing
member can use to add this client to the group asynchronously.A UserInitKey object specifies what ciphersuites a client supports,
as well as providing public keys that the client can use for key
derivation and signing. The client’s identity key is intended to be
stable throughout the lifetime of the group; there is no mechanism to
change it. Init keys are intended to be used a very limited number of
times, potentially once. (see ). UserInitKeys
also contain an identifier chosen by the client, which the client
MUST assure uniquely identifies a given UserInitKey object among the
set of UserInitKeys created by this client.The init_keys array MUST have the same length as the cipher_suites
array, and each entry in the init_keys array MUST be a public key
for the asymmetric encryption scheme defined in the cipher_suites array
and used in the HPKE construction for TreeKEM.The whole structure is signed using the client’s identity key. A
UserInitKey object with an invalid signature field MUST be
considered malformed. The input to the signature computation
comprises all of the fields except for the signature field.Over the lifetime of a group, its state will change for:Group initializationA current member adding a new clientA current member updating its leaf keyA current member deleting another current memberIn MLS, these changes are accomplished by broadcasting “handshake”
messages to the group. Note that unlike TLS and DTLS, there is not
a consolidated handshake phase to the protocol. Rather, handshake
messages are exchanged throughout the lifetime of a group, whenever
a change is made to the group state. This means an unbounded number
of interleaved application and handshake messages.An MLS handshake message encapsulates a specific “key exchange” message that
accomplishes a change to the group state. It also includes a
signature by the sender of the message over the GroupState object
representing the state of the group after the change has been made.The high-level flow for processing a Handshake message is as
follows:Verify that the prior_epoch field of the Handshake message
is equal the epoch field of the current GroupState object.Use the operation message to produce an updated, provisional
GroupState object incorporating the proposed changes.Look up the public key for slot index signer_index from the
roster in the current GroupState object (before the update).Use that public key to verify the signature field in the
Handshake message, with the updated GroupState object as input.If the signature fails to verify, discard the updated GroupState
object and consider the Handshake message invalid.Use the confirmation_key for the new group state to
compute the confirmation MAC for this message, as described below,
and verify that it is the same as the confirmation field.If the the above checks are successful, consider the updated
GroupState object as the current state of the group.The signature and confirmation values are computed over the
transcript of group operations, using the transcript hash from the
provisional GroupState object:[[ OPEN ISSUE: The confirmation data and signature data should probably
cover the same data as the one we cover with the GroupState. ]]HMAC uses the Hash algorithm for the ciphersuite in
use. Sign uses the signature algorithm indicated by the signer’s
credential in the roster.[[ OPEN ISSUE: The Add and Remove operations create a “double-join”
situation, where a member’s leaf key is also known to another
client. When a member A is double-joined to another B,
deleting A will not remove them from the conversation, since they
will still hold the leaf key for B. These situations are resolved
by updates, but since operations are asynchronous and members
may be offline for a long time, the group will need to be able to
maintain security in the presence of double-joins. ]][[ OPEN ISSUE: It is not possible for the recipient of an handshake
message to verify that ratchet tree information in the message is
accurate, because each node can only compute the secret and private
key for nodes in its direct path. This creates the possibility
that a malicious participant could cause a denial of service by sending
a handshake message with invalid values in the ratchet tree. ]][[ OPEN ISSUE: Direct initialization is currently undefined. A client can
create a group by initializing its own state to reflect a group
including only itself, then adding the initial members. This
has computation and communication complexity O(N log N) instead of
the O(N) complexity of direct initialization. ]]In order to add a new member to the group, an existing member of the
group must take two actions:Send a Welcome message to the new memberSend an Add message to the group (including the new member)The Welcome message contains the information that the new member
needs to initialize a GroupState object that can be updated to the
current state using the Add message. This information is encrypted
for the new member using HPKE. The recipient key pair for the
HPKE encryption is the one included in the indicated UserInitKey,
corresponding to the indicated ciphersuite.Note that the init_secret in the Welcome message is the
init_secret at the output of the key schedule diagram in
. That is, if the epoch value in the Welcome
message is n, then the init_secret value is init_secret_[n].
The new member can combine this init secret with the update secret
transmitted in the corresponding Add message to get the epoch secret
for the epoch in which it is added. No secrets from prior epochs
are revealed to the new member.Since the new member is expected to process the Add message for
itself, the Welcome message should reflect the state of the group
before the new user is added. The sender of the Welcome message can
simply copy all fields from their GroupState object.[[ OPEN ISSUE: The Welcome message needs to be synchronized in the
same way as the Add. That is, the Welcome should be sent only if
the Add succeeds, and is not in conflict with another, simultaneous
Add. ]]An Add message provides existing group members with the information
they need to update their GroupState with information about the new
member:The index field indicates where in the tree the new member should
be added. The new member can be added at an existing, blank leaf
node, or at the right edge of the tree. In any case, the index
value MUST satisfy 0 <= index <= n, where n is the size of the
group. The case index = n indicates an add at the right edge of
the tree). If index < n and the leaf node at position index is
not blank, then the recipient MUST reject the Add as malformed.The welcome_info_hash field contains a hash of the WelcomeInfo
object sent in a Welcome message to the new member.A group member generates this message by requesting a UserInitKey
from the directory for the user to be added, and encoding it into an
Add message.The client joining the group processes Welcome and Add
messages together as follows:Prepare a new GroupState object based on the Welcome messageProcess the Add message as an existing member wouldAn existing member receiving a Add message first verifies
the signature on the message, then updates its state as follows:If the index value is equal to the size of the group, increment
the size of the group, and extend the tree and roster accordinglyVerify the signature on the included UserInitKey; if the signature
verification fails, abortGenerate a WelcomeInfo object describing the state prior to the
add, and verify that its hash is the same as the value of the
welcome_info_hash fieldSet the roster entry at position index to the credential in the
included UserInitKeyUpdate the ratchet tree by setting to blank all nodes in the
direct path of the new nodeSet the leaf node in the tree at position index to a new node
containing the public key from the UserInitKey in the Add
corresponding to the ciphersuite in useThe update secret resulting from this change is an all-zero octet
string of length Hash.length.After processing an Add message, the new member SHOULD send an Update
immediately to update its key. This will help to limit the tree structure
degrading into subtrees, and thus maintain the protocol’s efficiency.An Update message is sent by a group member to update its leaf
secret and key pair. This operation provides post-compromise security with
regard to the member’s prior leaf private key.The sender of an Update message creates it in the following way:Generate a fresh leaf key pairCompute its direct path in the current ratchet treeA member receiving a Update message first verifies
the signature on the message, then updates its state as follows:Update the cached ratchet tree by replacing nodes in the direct
path from the updated leaf using the information contained in the
Update messageThe update secret resulting from this change is the secret for the
root node of the ratchet tree.A Remove message is sent by a group member to remove one or more
members from the group.The sender of a Remove message generates it as as follows:Generate a fresh leaf key pairCompute its direct path in the current ratchet tree, starting from
the removed leafA member receiving a Remove message first verifies
the signature on the message. The member then updates its
state as follows:Update the roster by setting the credential in the removed slot to
the null optional valueUpdate the ratchet tree by replacing nodes in the direct
path from the removed leaf using the information in the Remove messageReduce the size of the roster and the tree until the rightmost
element roster element and leaf node are non-nullUpdate the ratchet tree by setting to blank all nodes in the
direct path of the removed leafThe update secret resulting from this change is the secret for the
root node of the ratchet tree after the second step (after the third
step, the root is blank).[[ OPEN ISSUE: This section has an initial set of considerations
regarding sequencing. It would be good to have some more detailed
discussion, and hopefully have a mechanism to deal with this issue. ]]Each handshake message is premised on a given starting state,
indicated in its prior_epoch field. If the changes implied by a
handshake messages are made starting from a different state, the
results will be incorrect.This need for sequencing is not a problem as long as each time a
group member sends a handshake message, it is based on the most
current state of the group. In practice, however, there is a risk
that two members will generate handshake messages simultaneously,
based on the same state.When this happens, there is a need for the members of the group to
deconflict the simultaneous handshake messages. There are two
general approaches:Have the delivery service enforce a total orderHave a signal in the message that clients can use to break tiesAs long as handshake messages cannot be merged, there is a risk of starvation. In a sufficiently
busy group, a given member may never be able to send a handshake
message, because he always loses to other members. The degree to
which this is a practical problem will depend on the dynamics of the
application.It might be possible, because of the non-contributivity of intermediate nodes,
that update messages could be applied one after the other without the Delivery
Service having to reject any handshake message, which would make MLS
more resilient regarding the concurrency of handshake messages.
The Messaging system can decide to choose the order for applying
the state changes. Note that there are certain cases (if no total ordering
is applied by the Delivery Service) where the ordering is important
for security, ie. all updates must be executed before removes.Regardless of how messages are kept in sequence, implementations
MUST only update their cryptographic state when valid handshake messages
are received. Generation of handshake messages MUST be stateless,
since the endpoint cannot know at that time whether the change
implied by the handshake message will succeed or not.With this approach, the delivery service ensures that incoming messages are added to an
ordered queue and outgoing messages are dispatched in the same order. The server
is trusted to resolve conflicts during race-conditions (when two members send a
message at the same time), as the server doesn’t have any additional knowledge
thanks to the confidentiality of the messages.Messages should have a counter field sent in clear-text that can be checked by
the server and used for tie-breaking. The counter starts at 0 and is incremented
for every new incoming message. If two group members send a message with the same
counter, the first message to arrive will be accepted by the server and the second
one will be rejected. The rejected message needs to be sent again with the correct
counter number.To prevent counter manipulation by the server, the counter’s integrity can be
ensured by including the counter in a signed message envelope.This applies to all messages, not only state changing messages.Order enforcement can be implemented on the client as well, one way to achieve it
is to use a two step update protocol: the first client sends a proposal to update and
the proposal is accepted when it gets 50%+ approval from the rest of the group,
then it sends the approved update. Clients which didn’t get their proposal accepted,
will wait for the winner to send their update before retrying new proposals.While this seems safer as it doesn’t rely on the server, it is more complex and
harder to implement. It also could cause starvation for some clients if they keep
failing to get their proposal accepted.It is possible in principle to partly address the problem
of concurrent changes by having the recipients of the changes merge
them, rather than having the senders retry. Because the value of
intermediate node is determined by its last updated child,
updates can be merged
by recipients as long as the recipients agree on an order – the
only question is which node was last updated.Recall that the processing of an update proceeds in two steps:Compute updated secret values by hashing up the treeUpdate the tree with the new secret and public valuesTo merge an ordered list of updates, a recipient simply performs
these updates in the specified order.For example, suppose we have a tree in the following configuration:Now suppose B and C simultaneously decide to update to X and Y,
respectively. They will send out updates of the following form:Assuming that the ordering agreed by the group says that B’s update
should be processed before C’s, the other members in the group
will overwrite the root value for B with the root value from C, and
all arrive at the following state:The primary purpose of the handshake protocol is to provide an authenticated
group key exchange to clients. In order to protect Application messages
sent among those members of a group, the Application secret provided by the Handshake
key schedule is used to derive encryption keys for the Message Protection Layer.Application messages MUST be protected with the Authenticated-Encryption
with Associated-Data (AEAD) encryption scheme associated with the MLS ciphersuite.
Note that “Authenticated” in this context does not mean messages are known to
be sent by a specific client but only from a legitimate member of the group.
To authenticate a message from a particular member, signatures are required.
Handshake messages MUST use asymmetric signatures to strongly authenticate
the sender of a message.Each member maintains their own chain of Application secrets, where the first
one is derived based on a secret chained from the Epoch secret.
As shown in , the initial Application secret is bound to the
identity of each client to avoid collisions and allow support for decryption
of reordered messages.Subsequent Application secrets MUST be rotated for each message sent in
order to provide stronger cryptographic security guarantees. The Application
Key Schedule use this rotation to generate fresh AEAD encryption keys and nonces
used to encrypt and decrypt future Application messages.
In all cases, a participant MUST NOT encrypt more than expected by the security
bounds of the AEAD scheme used.Note that each change to the Group through a Handshake message will cause
a change of the group Secret. Hence this change MUST be applied before encrypting
any new Application message. This is required for confidentiality reasons
in order for members to avoid receiving messages from the group after leaving,
being added to, or excluded from the group.After computing the initial group Application Secret, which is derived from the
main key schedule, each member creates an initial sender Application Secret
to be used for its own sending chain:Note that [sender] represents the index of the member in the roster.Updating the Application secret and deriving the associated AEAD key and nonce can
be summarized as the following Application key schedule where
each member’s Application secret chain looks as follows after the initial
derivation:The Application context provided together with the previous Application secret
is used to bind the Application messages with the next key and add some freshness.[[OPEN ISSUE: The HKDF context field is left empty for now.
A proper security study is needed to make sure that we do not need
more information in the context to achieve the security goals.]][[ OPEN ISSUE: At the moment there is no contributivity of Application secrets
chained from the initial one to the next generation of Epoch secret. While this
seems safe because cryptographic operations using the application secrets can’t
affect the group init_secret, it remains to be proven correct. ]]The following rules apply to an Application Secret:Senders MUST only use the Application Secret once and monotonically
increment the generation of their secret. This is important to provide
Forward Secrecy at the level of Application messages. An attacker getting
hold of a member specific Application Secret at generation [N+1] will not be
able to derive the member’s Application Secret [N] nor the associated
AEAD key and nonce.Receivers MUST delete an Application Secret once it has been used to
derive the corresponding AEAD key and nonce as well as the next Application
Secret. Receivers MAY keep the AEAD key and nonce around for some
reasonable period.Receivers MUST delete AEAD keys and nonces once they have been used to
successfully decrypt a message.The Application AEAD keying material is generated from the following
input values:The Application Secret value;A purpose value indicating the specific value being generated;The length of the key being generated.Note, that because the identity of the participant using the keys to send data
is included in the initial Application Secret, all successive updates to the
Application secret will implicitly inherit this ownership.All the traffic keying material is recomputed whenever the underlying
Application Secret changes.The group members MUST use the AEAD algorithm associated with
the negotiated MLS ciphersuite to AEAD encrypt and decrypt their
Application messages and sign them as follows:The group identifier and epoch allow a device to know which group secrets
should be used and from which Epoch secret to start computing other secrets
and keys. The sender identifier is used to derive the member’s
Application secret chain from the initial group Application secret.
The application generation field is used to determine which Application
secret should be used from the chain to compute the correct AEAD keys
before performing decryption.The signature field allows strong authentication of messages:The signature used in the ApplicationMessageContent is computed over the SignatureContent
which covers the metadata information about the current state
of the group (group identifier, epoch, generation and sender’s Leaf index)
to prevent group members from impersonating other clients. It is also
necessary in order to prevent cross-group attacks.Application messages SHOULD be padded to provide some resistance
against traffic analysis techniques over encrypted traffic.
While MLS might deliver the same payload less frequently across
a lot of ciphertexts than traditional web servers, it might still provide
the attacker enough information to mount an attack. If Alice asks Bob:
“When are we going to the movie ?” the answer “Wednesday” might be leaked
to an adversary by the ciphertext length. An attacker expecting Alice to
answer Bob with a day of the week might find out the plaintext by
correlation between the question and the length.Similarly to TLS 1.3, if padding is used, the MLS messages MUST be
padded with zero-valued bytes before AEAD encryption. Upon AEAD decryption,
the length field of the plaintext is used to compute the number of bytes
to be removed from the plaintext to get the correct data.
As the padding mechanism is used to improve protection against traffic
analysis, removal of the padding SHOULD be implemented in a “constant-time”
manner at the MLS layer and above layers to prevent timing side-channels that
would provide attackers with information on the size of the plaintext.
The padding length length_of_padding can be chosen at the time of the message
encryption by the sender. Recipients can calculate the padding size from knowing
the total size of the ApplicationPlaintext and the length of the content.[[ TODO: A preliminary formal security analysis has yet to be performed on
this authentication scheme.]][[ OPEN ISSUE: Currently, the group identifier, epoch and generation are
contained as meta-data of the Signature. A different solution could be to
include the GroupState instead, if more information is required to achieve
the security goals regarding cross-group attacks. ]][[ OPEN ISSUE: Should the padding be required for Handshake messages ?
Can an adversary get more than the position of a participant in the tree
without padding ? Should the base ciphertext block length be negotiated or
is is reasonable to allow to leak a range for the length of the plaintext
by allowing to send a variable number of ciphertext blocks ? ]]Since each Application message contains the group identifier, the epoch and a
message counter, a client can receive messages out of order.
If they are able to retrieve or recompute the correct AEAD decryption key
from currently stored cryptographic material clients can decrypt
these messages.For usability, MLS clients might be required to keep the AEAD key
and nonce for a certain amount of time to retain the ability to decrypt
delayed or out of order messages, possibly still in transit while a
decryption is being done.[[TODO: Describe here or in the Architecture spec the details. Depending
on which Secret or key is kept alive, the security guarantees will vary.]]The security goals of MLS are described in [I-D.ietf-mls-architecture]. We describe here how the
protocol achieves its goals at a high level, though a complete security analysis is outside of the
scope of this document.Group secrets are derived from (i) previous group secrets, and (ii) the root key of a ratcheting
tree. Only group members know their leaf private key in the group, therefore, the root key of the
group’s ratcheting tree is secret and thus so are all values derived from it.Initial leaf keys are known only by their owner and the group creator, because they are derived from
an authenticated key exchange protocol. Subsequent leaf keys are known only by their owner. [[TODO:
or by someone who replaced them.]]Note that the long-term identity keys used by the protocol MUST be distributed by an “honest”
authentication service for clients to authenticate their legitimate peers.There are two forms of authentication we consider. The first form
considers authentication with respect to the group. That is, the group
members can verify that a message originated from one of the members
of the group. This is implicitly guaranteed by the secrecy of the
shared key derived from the ratcheting trees: if all members of the
group are honest, then the shared group key is only known to the group
members. By using AEAD or appropriate MAC with this shared key, we can
guarantee that a member in the group (who knows the shared secret
key) has sent a message.The second form considers authentication with respect to the sender,
meaning the group members can verify that a message originated from a
particular member of the group. This property is provided by digital
signatures on the messages under identity keys.[[ OPEN ISSUE: Signatures under the identity keys, while simple, have
the side-effect of preclude deniability. We may wish to allow other options, such as (ii) a key
chained off of the identity key, or (iii) some other key obtained
through a different manner, such as a pairwise channel that
provides deniability for the message contents.]]Message encryption keys are derived via a hash ratchet, which provides a form of forward secrecy: learning a
message key does not reveal previous message or root keys. Post-compromise security is provided by
Update operations, in which a new root key is generated from the latest ratcheting tree. If the
adversary cannot derive the updated root key after an Update operation, it cannot compute any
derived secrets.Initialization keys are intended to be used only once and then deleted. Reuse of init keys is not believed to be
inherently insecure , although it can complicate protocol analyses.TODO: Registries for protocol parameters, e.g., ciphersuitesBenjamin Beurdouche
INRIA
benjamin.beurdouche@ens.frKarthikeyan Bhargavan
INRIA
karthikeyan.bhargavan@inria.frCas Cremers
University of Oxford
cas.cremers@cs.ox.ac.ukAlan Duric
Wire
alan@wire.comSrinivas Inguva
Twitter
singuva@twitter.comAlbert Kwon
MIT
kwonal@mit.eduEric Rescorla
Mozilla
ekr@rtfm.comThyla van der Merwe
Royal Holloway, University of London
thyla.van.der@merwe.techPublic Key Cryptography For The Financial Services Industry: The Elliptic Curve Digital Signature Algorithm (ECDSA)ANSIIEEE Standard Specifications for Password-Based Public-Key Cryptographic TechniquesKey words for use in RFCs to Indicate Requirement LevelsIn many standards track documents several words are used to signify the requirements in the specification. These words are often capitalized. This document defines these words as they should be interpreted in IETF documents. This document specifies an Internet Best Current Practices for the Internet Community, and requests discussion and suggestions for improvements.Ambiguity of Uppercase vs Lowercase in RFC 2119 Key WordsRFC 2119 specifies common key words that may be used in protocol specifications. This document aims to reduce the ambiguity by clarifying that only UPPERCASE usage of the key words have the defined special meanings.The Transport Layer Security (TLS) Protocol Version 1.3This document specifies version 1.3 of the Transport Layer Security (TLS) protocol. TLS allows client/server applications to communicate over the Internet in a way that is designed to prevent eavesdropping, tampering, and message forgery.This document updates RFCs 5705 and 6066, and obsoletes RFCs 5077, 5246, and 6961. This document also specifies new requirements for TLS 1.2 implementations.An Interface and Algorithms for Authenticated EncryptionThis document defines algorithms for Authenticated Encryption with Associated Data (AEAD), and defines a uniform interface and a registry for such algorithms. The interface and registry can be used as an application-independent set of cryptoalgorithm suites. This approach provides advantages in efficiency and security, and promotes the reuse of crypto implementations. [STANDARDS-TRACK]Elliptic Curves for SecurityThis memo specifies two elliptic curves over prime fields that offer a high level of practical security in cryptographic applications, including Transport Layer Security (TLS). These curves are intended to operate at the ~128-bit and ~224-bit security level, respectively, and are generated deterministically based on a list of required properties.Hybrid Public Key EncryptionThis document describes a scheme for hybrid public-key encryption (HPKE). This scheme provides authenticated public key encryption of arbitrary-sized plaintexts for a recipient public key. HPKE works for any Diffie-Hellman group and has a strong security proof. We provide instantiations of the scheme using standard and efficient primitives.HMAC-based Extract-and-Expand Key Derivation Function (HKDF)This document specifies a simple Hashed Message Authentication Code (HMAC)-based key derivation function (HKDF), which can be used as a building block in various protocols and applications. The key derivation function (KDF) is intended to support a wide range of applications and requirements, and is conservative in its use of cryptographic hash functions. This document is not an Internet Standards Track specification; it is published for informational purposes.HMAC: Keyed-Hashing for Message AuthenticationThis document describes HMAC, a mechanism for message authentication using cryptographic hash functions. HMAC can be used with any iterative cryptographic hash function, e.g., MD5, SHA-1, in combination with a secret shared key. The cryptographic strength of HMAC depends on the properties of the underlying hash function. This memo provides information for the Internet community. This memo does not specify an Internet standard of any kindOn Ends-to-Ends Encryption: Asynchronous Group Messaging with Strong Security GuaranteesA Formal Security Analysis of the Signal Messaging ProtocolOn reusing ephemeral keys in Diffie-Hellman key agreement protocolsRecommendation for Pair-Wise Key Establishment Schemes Using Discrete Logarithm CryptographyThe Double Ratchet AlgorithmCertificate Transparency Version 2.0This document describes version 2.0 of the Certificate Transparency (CT) protocol for publicly logging the existence of Transport Layer Security (TLS) server certificates as they are issued or observed, in a manner that allows anyone to audit certification authority (CA) activity and notice the issuance of suspect certificates as well as to audit the certificate logs themselves. The intent is that eventually clients would refuse to honor certificates that do not appear in a log, effectively forcing CAs to add all issued certificates to the logs. This document obsoletes RFC 6962. It also specifies a new TLS extension that is used to send various CT log artifacts. Logs are network services that implement the protocol operations for submissions and queries that are defined in this document.I Know Why You Went to the Clinic: Risks and Realization of HTTPS Traffic AnalysisHTTPS traffic analysis and client identification using passive SSL/TLS fingerprintingOne benefit of using left-balanced trees is that they admit a simple
flat array representation. In this representation, leaf nodes are
even-numbered nodes, with the n-th leaf at 2*n. Intermediate nodes
are held in odd-numbered nodes. For example, a 11-element tree has
the following structure:This allows us to compute relationships between tree nodes simply by
manipulating indices, rather than having to maintain complicated
structures in memory, even for partial trees. The basic
rule is that the high-order bits of parent and child nodes have the
following relation (where x is an arbitrary bit string):The following python code demonstrates the tree computations
necessary for MLS. Test vectors can be derived from the diagram
above.