Archive reconstruction: This is an independently written research summary for a historical URL. It is not the original article, and this site is not affiliated with the original author.
A case study in how an anonymized social graph can disclose hidden relationships when compared with external network data.
Historical context
The research examined graph matching as a way to identify nodes in a dataset whose personal labels had been obscured. Relationships between nodes can act as identifying evidence even when familiar fields such as usernames are removed.
Why the topic matters
Machine-learning challenges that publish stripped-down network data need to consider information leakage from datasets outside the competition. A model may appear to predict unseen connections while actually discovering a correspondence to publicly available records.
Reading the material today
This archival annotation preserves the paper’s historical place in debates about contest design, data release and re-identification. It is not a new claim about any current competition or a reproduction of the original article.
Historical source: Read the original material preserved at 33bits.wordpress.com.
