Locating the Source in Real-world Diffusion Network

TL;DR

This paper introduces a practical and efficient method for identifying the original source of information spread in real-world networks, significantly improving accuracy and speed over previous approaches.

Contribution

It presents a new source detection method that handles realistic scenarios, achieves higher accuracy, and is computationally faster than existing techniques.

Findings

Top ten accuracy improved to ~30% from <10%

Method is about 10 times faster than previous work

Effective in real-world propagation networks

Abstract

The problem of identifying the source of a propagation based on limited observations has been studied significantly in recent years, as it can help reducing the damage caused by unwanted infections. In this paper we present an efficient approach to find the node that originally introduced a piece of information into the network, and infer the time when it is initiated. Labeling infected nodes detected in limited observation as observed nodes and other ones as hidden nodes, we first estimate the shortest path between hidden nodes to observed ones for each propagation trace. Then we find the best node as the source among the hidden nodes by optimizing over square loss function. The method presented in this paper is based on more realistic situations and is easy and more practical than previous works. Our experiments on real-world propagation through networks show the superiority of our approach in detecting true source, boosting the top ten accuracy from less than 10% for the sate-of-the-art methods to approximately 30%. Additionally, we observe that our source identification method runs about 10 times faster than previous work.