Briar: Peer Moderation

Briar uses a mechanism called peer moderation to limit the impact of automated spam. In peer moderation, each subscriber keeps ratings for pseudonyms that have previously posted messages. A rating has one of three values: negative, meaning that the subscriber considers the pseudonym to be a spammer; neutral, meaning that the subscriber has not decided whether the pseudonym is a spammer; or positive, meaning that the subscriber does not consider the pseudonym to be a spammer. All previously unknown pseudonyms start with neutral ratings.

When a pseudonymous message is posted to an unrestricted group, each subscriber who receives the message checks whether she has a positive rating for the pseudonym that signed the message. If so, the subscriber forwards the message to any contacts who are known to subscribe to the group. If not, the subscriber does not forward the message immediately, but may forward it later if she assigns a positive rating to the pseudonym.

Anonymous messages may also be rated, but each such rating only applies to a single message.

Peer moderation may cause a group's subscribers to see different sets of messages. To minimise the disruption this could cause to the flow of conversation, Briar uses a rule called backward inclusion to ensure that each subscriber sees complete conversations. The backward inclusion rule specifies that if a message is eligible for forwarding then all its ancestors in the same thread are also eligible for forwarding.

A message posted to an unrestricted group may therefore reach a wide audience if either (a) it is signed by a pseudonym for which many subscribers have positive ratings, (b) it is an anonymous message to which many subscribers have given positive ratings, or (c) any of its descendants in the same thread meets either of the first two conditions.

There is nothing to stop a spammer from using any number of pseudonyms to sign her messages, but unless other subscribers create positive ratings for those pseudonyms, the spammer's messages will only reach her own contacts.

New users who are not spammers will not initially be able to reach large audiences with their messages, but once they have received some positive ratings their messages will start to reach more subscribers.

Backward inclusion ensures that if a pseudonym that is positively rated by many subscribers replies to a message by an unknown pseudonym, the unknown pseudonym's message will reach the same audience as the reply. This allows new pseudonyms to be discovered, but makes it important not to reply to spam, since doing so may help to distribute the spam.