Time-Efficient Network Monitoring Through Confined Search and Adaptive Evaluation

Abstract

The network monitoring problem, crucial to many applications from outbreak prevention to online rumor management, demands an optimal set of monitors to detect the spreading of infections or rumors over a network. We tackle this problem through solving a type of facility location problem where the monitored nodes are selected to minimize their distance to other nodes. Existing methods for this problem either consume prohibitively long time for large networks, lack of reasonable theoretical performance guarantees, or are very difficult to implement. We propose a new algorithm, $\mathsf {csav}$, which combines a novel technique to reduce the search space with an iterative improvement mechanism. Our algorithm outputs a logarithmic number of monitors in $\tilde{O}(|E|)$ time. We perform empirical analysis over both synthesized and real-world networks as well as three propagation models. The results show that $\mathsf {csav}$ achieves superior performance over a number of benchmark algorithms. In particular, it produces outputs that are comparable to the well-established local search at only a fraction of its running time. Our approach is hence a scalable and time-efficient method for the network monitoring problem.

This work was supported by the National Natural Science Foundation of China (61772503).