Analyzing Social Network Metadata to Uncover Censorship
BY Jessica McKenzie | Wednesday, March 12 2014
If you've entered your email into the MIT Media Lab Immersion platform, you might have some idea of the information that can be gleaned from metadata. The same is true of social networks like Twitter and Facebook. One researcher has found that analysis of social network metadata can reveal wide scale censorship with 85 percent accuracy, without needing to track sensitive keywords.
Donn Morrison at the Norwegian University of Science and Technology in Trondheim created a computer-simulated social network, where two users were regarded as being connected if one of their posts appeared in the other's timeline. The pattern of connections between users provided the metadata that he used to analyse network behaviour. [sic]
Most social networks are made up of clusters of communities, the links between them creating a characteristic structure. But when Morrison simulated the actions of state censors who deleted at least 10 percent of posts, the missing links changed the shape of the entire network, leaving it malformed and less connected. This was especially true when the censors targeted popular posts that had been retweeted.
Using metadata to reveal censorship is easier to automate than tracking politically triggering words, which change over time. When applied to a real life social network, it could be programmed to alert users whose posts are altered or deleted.
Morrison hopes the program could be used to create a daily censorship report that one could check with the ease of checking the weather.
Personal Democracy Media is grateful to the Omidyar Network and the UN Foundation for their generous support of techPresident's WeGov section.