III: Small: Network Sampling and Construction Methods for Inference and Anonymization

University Of California-Irvine, Irvine CA

Investigators

Abstract

This project will study big network data, including but not limited to those generated by communication using mobile devices and online social media. The increasing availability of such datasets poses both opportunities and challenges. On one hand, their relatively easy collection and analysis facilitate the understanding of online human activity and the design of better services. This, however, requires that targeted answers can be efficiently provided based on limited (sampled or aggregated) data. On the other hand, the combination of big data and powerful inference techniques poses privacy concerns and an increasing need for effective anonymization techniques. This project will address both aspects and will advance the state of the art in network modeling and analysis. In the first part of the project, adaptive link-trace sampling will be designed for ERGM inference of network structure and/or attributes and for node-level analysis. Time series analysis will be applied to aggregate spatio-temporal network activity data to infer and predict patterns. In the second part of the project, novel models and algorithms will be designed in order to generate synthetic networks that resemble real ones with respect to target characteristics of interest. The methods will be applied to simulation and anonymization of network data and can improve the understanding of the tradeoff between utility and privacy in this domain. For further information see the project web site at: http://networkdata.calit2.uci.edu