Privacy preserving optimization of communication networks

Modern society takes connectivity for granted, relying heavily on communication networks, both for interpersonal connection and to support critical infrastructure. As Internet- and data-driven technologies become increasingly pervasive, our dependence on fast, reliable communication will only deepen, necessitating advanced tools for optimizing network efficiency and resilience. Such optimization must account for the interplay between the static network infrastructure and the dynamic user preferences. The challenge is that while the infrastructure data is accessible to network operators, the user preferences, tied to personal mobility and communication habits, are protected by privacy laws and are thus heavily restricted. To address this, we introduce CLUSTER: an interpretable Bayesian nonparametric framework that leverages aggregate, low-resolution, unprotected data to identify user groups with correlated connection patterns. By uncovering these patterns, we show, CLUSTER offers actionable insights, from scheduling base-station activation to guiding deployment of new stations - all without compromising user privacy. CLUSTER thus offers a principled approach to extract meaningful insights from restricted data. Modern communication networks must balance efficiency with user privacy. Here, authors develop a privacy-preserving interpretable Bayesian framework that enables network optimization using only aggregated data, without compromising individual user information.