Source Detection in Complex Networks

Source detection in complex networks seeks to identify the origin of a spreading process—whether an epidemic, a viral message or a cascading failure—using limited observations of the network’s dynamic state. Such networks are characterised by heterogeneous connectivity patterns, temporal variability and noise in transmission times. The core challenges lie in incomplete data, stochastic propagation and the potential presence of multiple sources. Contemporary approaches combine theoretical frameworks from graph controllability and statistical inference with computational techniques such as maximum‐likelihood estimation, compressive sensing and entropy‐based measures. Observer‐based methods exploit time‐stamped reports from selected nodes, while probabilistic strategies reconstruct source likelihoods via gradients or Bayesian updates. Recent advances address multiple simultaneous outbreaks, time‐varying topologies and strategic placement of measurement nodes. Practical applications span public‐health investigations, tracing misinformation on social platforms and diagnosing failures in critical infrastructures, underscoring the global importance of precise and efficient source localisation.

Research from Nature Portfolio

Recent studies have introduced entropy‐driven algorithms to locate multiple sources within an infected cluster. By defining two complementary entropy measures—one for the proportion of infected neighbours and one for the intensity of adjacent infections—researchers have developed a cohesion‐based cluster analysis that identifies multiple origins in both model and empirical networks with higher accuracy than previous heuristics. Complementary work has provided a systematic comparison of observer‐based localisation methods, evaluating how network topology, observer density and infection rate influence the precision of source estimates. This analysis highlights that correlation‐based techniques excel at low transmission rates, whereas multi‐path traversal methods become superior as contagion spreads more rapidly. Foundational efforts have also refined maximum‐likelihood algorithms by weighting high‐quality observer data to reduce computational complexity and enhance localisation accuracy in large‐scale, scale‐free networks.

Research from all publishers

A general theoretical framework leveraging network controllability combined with compressive sensing has been formulated to achieve optimal source localisation from minimal observations. By identifying a minimal set of “messenger” nodes whose outputs provide sufficient information, the problem reduces to sparse signal reconstruction, enabling reliable detection even with a single messenger in certain undirected networks. Studies based on compartmental epidemic models on social platforms have adopted susceptible-exposed-infected-recovered dynamics to derive an estimator that maximises the probability of matching the true source from a single snapshot, demonstrating superior performance over centrality‐based heuristics across synthetic and real-world networks. More recently, methods integrating diffusion direction information with observer timing—using a relaxed direction-induced search and complementary similarity measures—have shown improved source localisation by reconstructing approximate diffusion trees and comparing both the order and interval of reported infection times.

Source Detection in Complex Networks publication trend

The graph below shows the total number of articles in source detection in complex networks across all publications each year (not limited to Nature Index journals).

Technical terms

Complex network: A graph whose nodes and edges exhibit non‐trivial patterns of connectivity, often with heterogeneous degree distributions and clustering.

Observer: A node selected to report its infection time or state, serving as a data point for source estimation.

Messenger node: A minimal set of observers whose collective measurements suffice to reconstruct the source location via sparse inference.

Compressive sensing: A signal‐processing technique that recovers sparse signals from underdetermined measurements by exploiting signal sparsity.

Neighbourhood entropy: A quantitative measure of infection information in a node’s local neighbourhood, encompassing both the adjacency of infected neighbours and the intensity of their infections.

Direction‐induced search: An approach that incorporates diffusion directionality, inferred from observer reports, to approximate the actual propagation tree for source localisation.

References

1. Fast and accurate detection of spread source in large complex networks. Scientific Reports (2018).
2. Optimal localization of diffusion sources in complex networks. Royal Society Open Science (2017).
3. Locating multiple diffusion sources in time varying networks from sparse observations. Scientific Reports (2018).
4. Rumor Source Detection in Networks Based on the SEIR Model. IEEE Access (2019).
5. The effect of transmission variance on observer placement for source-localization. Applied Network Science (2017).
6. Locating the propagation source in complex networks with observers-based similarity measures and direction-induced search. Soft Computing (2023).
7. Multi-source detection based on neighborhood entropy in social networks. Scientific Reports (2022).
8. Comparison of observer based methods for source localisation in complex networks. Scientific Reports (2022).