A novel source locating strategy without consistent assumptions

Locating the source of propagation is a ubiquitous but challenging problem in the field of complex networks. The traditional source location methods based on a set of observers can achieve a high locating accuracy. However, such high accuracy is based on the consistent assumption which means the propagation delays consistently follow a certain distribution in both the infected time calculation process and the source location process. Based on our simulation results and existing researches, we find that the real propagation delays, in some real-world scenarios, often break such consistent assumption and the predication accuracy of existing methods decline significantly in these circumstances. Therefore it raises a critical question: can we locate the infection source without assuming the distribution of propagation delays? In this paper, we first formulate the problem of locating source as inferring the parameters of propagation delays based on a set of observers. Then, we propose a novel reverse propagation strategy to locate infection source. Finally, a comprehensive comparisons are used to provide a quantitative analyses of our method. The results show that our strategy has a higher accuracy than the traditional methods without the consistent assumptions.