Using Metrics Suites to Improve the Measurement of Privacy in Graphs

Social graphs are widely used in research (e.g., epidemiology) and business recommender systems). However, sharing these poses privacy risks because they contain sensitive information about individuals. Graph anonymization techniques aim to protect individual users a graph, while graph de-anonymization aims re-identify users. The effectiveness of algorithms is usually evaluated with metrics. it unclear how strong existing metrics when privacy. In this article, we study 26 for evaluate their strength terms three criteria: monotonicity indicates whether the metric lower stronger adversaries; within-scenario comparisons, evenness values spread evenly; between-scenario shared value range use consistent range across scenarios. Our extensive experiments indicate that no single fulfills all criteria perfectly. We therefore methods from multi-criteria decision analysis aggregate multiple suite, show suites improve monotonicity compared best metric. This important result enables more monotonic, thus accurate, evaluations new algorithms.