Facebook's map of the relationships between individual users is being brought to bear to filter out abusive and fake users from the social network.
The 500 million-strong social network is trialling a number of features to discourage people who bully others, post spam or contravene the site's policies in other ways, a manager for Facebook's public policy team told journalists on Tuesday.
When a user who typically only posts to seven or so walls a day is posting to 20 or 30, that indicates to us that the account has been compromised. – Simon Axten, Facebook
For example, Facebook is experimenting with actively banning members from parts of the site if they have been abusive. This includes measures such as "shutting [them] off from using [features] such as creating a group or maybe joining a group", Simon Axten said.
As part of this, the social-networking site uses the data it gleans from members' interaction with their friends to identify spammers.
"When a user who typically only posts to seven or so walls a day is posting to 20 or 30, that indicates to us that the account has been compromised," Axten said. "These systems look for behaviour that is anomalous and falls outside the norm of the typical Facebook user."
Axten said, in the future, the company plans to "use additional signals to the ones we have now to figure out whether [user] activity is bad".
Facebook has come under critcism in the past over the amount of data it collects on its members. Security expert Bruce Schneier and others have argued that this collection is detrimental to user privacy. In response, Facebook has said that its focus is on users' experience and not on selling data.
Fundamentally, much of the company's efforts to increase user comfort on the site revolves around encouraging people to use their own identities, rather than fake ones, according to Axten.
The use of real identities leads to an improvement in the quality of interaction on the site, he added. On 1 February, Facebook announced it is planning to launch a third-party commenting system for use on other web sites, and the effects could spread to these, Axten suggested.
"[We have received] anecdotal reports from website owners saying that allowing people to log in with their Facebook identity before making a comment or blog post on an article has forced much better behaviour," he said.