Mining Social Interactions in Privacy-preserving TemporalNetworks

The opportunities to empirically study temporal networks nowadays are immense thanks to Internet of Things technologies along with ubiquitous and pervasive computing that allow a real-time fine-grained collection of social network data. This empowers data analytics and data scientists to reason about complex temporal phenomena, such as disease spread, residential energy consumption, political conflicts etc., using systematic methodologies from complex networks and graph spectral analysis. However, a misuse of these methods may result in privacy-intrusive and discriminatory actions that may threaten citizens’ autonomy and put their life under surveillance. This paper studies highly sparse temporal networks that model social interactions such as the physical proximity of participants in conferences. When citizens can self-determine the anonymized proximity data they wish to share via privacy-preserving platforms, temporal networks may turn out to be highly sparse and have low quality. This paper shows that even in this challenging scenario of privacy-by-design, significant information can be mined from temporal networks such as the correlation of events happening during a conference or stable groups interacting over time. The findings of this paper contribute to the introduction of privacy-preserving data analytics intemporal networks and their applications.