Skip to content

Source localization

Identifying the origin or source node of information spread in networks

The problem of identifying the origin (source node) of information, rumors, or misinformation spreading through a network. Given partial observations of diffusion—limited information about which users have received a message and when—source localization aims to infer the source using network structure, temporal dynamics, and observer placement strategies.

Key concepts

Observer-based inference — Identifying sources from observations made by strategically placed monitors in the network. Observers record timing and direction of received information; the source is inferred by aggregating evidence across multiple observers and accounting for network paths.

Likelihood-based estimation — Source identification as a maximum-likelihood or Bayesian problem: given observed cascade data, which source node maximizes the likelihood of producing exactly those observations? Approaches differ in assumptions about propagation models, observer reliability, and computational tractability.

Structural credibility — Observer reliability varies by network position (high-degree nodes are more trustworthy; isolated nodes are noisier). Credibility weighting enables robust inference from heterogeneous observer networks.

Trade-offs in observer placement — Balancing coverage (fraction of network within observation range) against redundancy (multiple observers per candidate source for cross-validation). Optimization under resource constraints is NP-hard; metaheuristics like genetic algorithms and moth-flame optimization enable practical solutions.

Key papers

Papers on this topic