Protecting Privacy by Secure Computation
We consider collaborative social network analysis without revealing private inputs of the participants. This problem arises in criminal investigations of federal police organization where single organizations may not reveal their data without probable cause, but the aggregation of all data entails new information, such as the entire social network structure. We present algorithms for securely computing either the entire, anonymized graph or only specific metrics for individuals. We use secure computation protocols to disclose nothing, but the output of the analysis, i.e. anything that cannot be derived from one’s input and output – including other parties’ input – remains private. We have implemented a prototype for SAP’s investigative case management system – a derivate of its customer relationship management.