`11/WAKU2-RELAY` specifies a [Publish/Subscribe approach](https://docs.libp2p.io/concepts/publish-subscribe/) to peer-to-peer messaging with a strong focus on privacy, censorship-resistance, security and scalability.
Its current implementation is a minor extension of the [libp2p GossipSub protocol](https://github.com/libp2p/specs/blob/master/pubsub/gossipsub/README.md) and prescribes gossip-based dissemination.
As such the scope is limited to defining a separate [`protocol id`](https://github.com/libp2p/specs/blob/master/connections/README.md#protocol-negotiation) for `11/WAKU2-RELAY`, establishing privacy and security requirements, and defining how the underlying GossipSub is to be interpreted and implemented within the Waku and cryptoeconomic domain.
Note that data confidentiality, integrity, and authenticity are currently considered out of scope for `11/WAKU2-RELAY` and must be handled by higher layer protocols such as [`14/WAKU2-MESSAGE`](/spec/14).
<!-- May add the definition of the unsupported feature:
Confidentiality indicates that an adversary should not be able to learn the data carried by the `WakuRelay` protocol.
Integrity indicates that the data transferred by the `WakuRelay` protocol can not be tampered with by an adversarial entity without being detected.
Authenticity no adversary can forge data on behalf of a targeted publisher and make it accepted by other subscribers as if the origin is the target. -->
_Personally identifiable information_ (PII) refers to any piece of data that can be used to uniquely identify a user.
For example, the signature verification key, and the hash of one's static IP address are unique for each user and hence count as PII.
# Adversarial Model
- Any entity running the `11/WAKU2-RELAY` protocol is considered an adversary.
This includes publishers, subscribers, and all the peers' direct connections.
Furthermore, we consider the adversary as a passive entity that attempts to collect information from others to conduct an attack but it does so without violating protocol definitions and instructions.
For example, under the passive adversarial model, no malicious subscriber hides the messages it receives from other subscribers as it is against the description of `11/WAKU2-RELAY`.
However, a malicious subscriber may learn which topics are subscribed to by which peers.
The [PubSub interface specification](https://github.com/libp2p/specs/blob/master/pubsub/README.md) defines the protobuf RPC messages exchanged between peers participating in a GossipSub network.
We republish these messages here for ease of reference and define how `11/WAKU2-RELAY` uses and interprets each field.
The various [control messages](https://github.com/libp2p/specs/blob/master/pubsub/gossipsub/gossipsub-v1.0.md#control-messages) defined for GossipSub are used as specified there.
The [`StrictNoSign` option](https://github.com/libp2p/specs/blob/master/pubsub/README.md#signature-policy-options) MUST be used, to ensure that messages are built without the `signature`, `key`, `from` and `seqno` fields.
Note that this does not merely imply that these fields be empty, but that they MUST be _absent_ from the marshalled message.
<!-- TODO: realized that the prime security objective of the `WakuRelay` protocol is to provide peers unlinkability as such this feature is prioritized over other features e.g., unlinkability is preferred over authenticity and integrity. It might be good to motivate unlinkability and its impact on the relay protocol or other protocols invoking relay protocol.-->
To address publisher-message unlinkability, one should remove any PII from the published message.
As such, `11/WAKU2-RELAY` follows the `StrictNoSign` policy as described in [libp2p PubSub specs](https://github.com/libp2p/specs/tree/master/pubsub#message-signing).
As the result of the `StrictNoSign` policy, `Message`s should be built without the `from`, `signature` and `key` fields since each of these three fields individually counts as PII for the author of the message (one can link the creation of the message with libp2p peerId and thus indirectly with the IP address of the publisher).
Note that removing identifiable information from messages cannot lead to perfect unlinkability.
The direct connections of a publisher might be able to figure out which `Message`s belong to that publisher by analyzing its traffic.
The possibility of such inference may get higher when the `data` field is also not encrypted by the upper-level protocols. <!-- TODO: more investigation on traffic analysis attacks and their success probability-->
This allows an immediate subscriber-topic unlinkability where subscribers are not re-identifiable from their subscribed topic IDs as the entire network is linked to the same topic ID.
This level of unlinkability / anonymity is known as [k-anonymity](https://www.privitar.com/blog/k-anonymity-an-introduction/) where k is proportional to the system size (number of participants of Waku relay protocol).
However, note that `11/WAKU2-RELAY` supports the use of more than one topic.
In case that more than one topic id is utilized, preserving unlinkability is the responsibility of the upper-level protocols which MAY adopt [partitioned topics technique](https://specs.status.im/spec/10#partitioned-topic) to achieve K-anonymity for the subscribed peers.
In the spam-protected `11/WAKU2-RELAY` protocol, no adversary can flood the system with spam messages (i.e., publishing a large number of messages in a short amount of time).
Spam protection is partly provided by GossipSub v1.1 through [scoring mechanism](https://github.com/libp2p/specs/blob/master/pubsub/gossipsub/gossipsub-v1.1.md#spam-protection-measures).
At a high level, peers utilize a scoring function to locally score the behavior of their connections and remove peers with a low score.
`11/WAKU2-RELAY` aims at enabling an advanced spam protection mechanism with economic disincentives by utilizing Rate Limiting Nullifiers.
In a nutshell, peers must conform to a certain message publishing rate per a system-defined epoch, otherwise, they get financially penalized for exceeding the rate.
- Providing **Unlinkability**, **Integrity** and **Authenticity** simultaneously:
Integrity and authenticity are typically addressed through digital signatures and Message Authentication Code (MAC) schemes, however, the usage of digital signatures (where each signature is bound to a particular peer) contradicts with the unlinkability requirement (messages signed under a certain signature key are verifiable by a verification key that is bound to a particular publisher).
As such, integrity and authenticity are missing features in `11/WAKU2-RELAY` in the interest of unlinkability.
In future work, advanced signature schemes like group signatures can be utilized to enable authenticity, integrity, and unlinkability simultaneously.
In a group signature scheme, a member of a group can anonymously sign a message on behalf of the group as such the true signer is indistinguishable from other group members. <!-- TODO: shall I add a reference for group signatures?-->
