Contrasting the Spread of Misinformation in Online Social Networks

Authors: Marco Amoruso, Daniele Anello, Vincenzo Auletta, Raffaele Cerulli, Diodato Ferraioli, Andrea Raiconi

JAIR 2020 | Venue PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental To verify the quality and efficiency of our suggested solutions, we conduct experiments on several real-world networks. The results of this extensive experimental phase validate our heuristics as effective tools to contrast the spread of misinformation in online social networks.
Researcher Affiliation Academia Marco Amoruso marco.amorusogmail.com Daniele Anello EMAIL Vincenzo Auletta EMAIL Raffaele Cerulli EMAIL Diodato Ferraioli EMAIL Andrea Raiconi EMAIL University of Salerno Via Giovanni Paolo II 132, 84084 Fisciano (SA), Italy
Pseudocode Yes Summarizing, our procedure works as described in Algorithm 1. Input: Graph G, Sources S, Targets T, and integer k. Output: Monitor vertices M. 1 G , s , t = Source Contraction(G, S, T) 2 ˆG = Weight Conversion(G ) 3 (L, R) = Unbalanced Cut( ˆG, s , t , k) 4 C = Bipartite Graph From Cut(L, R) 5 M = Vertex Cover(C) 6 return M Algorithm 1: Algorithm for monitoring placement
Open Source Code No The algorithms and the Independent Cascade infection model are implemented in Python. However, the paper does not provide an explicit link or statement about open-sourcing the code for the methodology described.
Open Datasets Yes We validated the proposed approaches by performing experiments on a set of freely available benchmark instances, deriving from real-world data. ... we considered the following 8 instances coming from the Social category of the Konect database1: Advogato (from now, referred to as Advogato); Digg Friends (Digg), Epinions trust (Epinions-2), Google+ (Gplus), Slashdot Zoo (Slashdot), Twitter lists (Twitter), Youtube friendship (Youtube-1), Youtube links (Youtube2). Finally, we considered the Political blogs (Polblogs) network of hyperlinks, available through the database of network data maintained by Mark Newman2. ... The two considered undirected graphs are Facebook (NIPS) (Facebook-1) and Facebook friendships (Facebook-2), again available through the Konect database. ... we considered also two smaller networks, namely Wiki-Vote and Gnutella08 (Leskovec & Krevl, 2014)
Dataset Splits Yes We chose 2,3 and 5 as values for |S|, and simulated the process for 3 different random choices of the sources for each value of |S| (for a total of 9 simulations for each network). ... For each of these experimental settings, i.e., for every graph, each value of k, and each range, the experiment has been repeated at least 15 times. ... To force each test to be in one of these ranges, we choose a random integer i within that range, and we stop the cascade process as soon as i nodes have been infected by misinformation.
Hardware Specification Yes All tests were conducted on a machine running Cent OS Linux 7, equipped with an Intel Xeon E5-2650 v3 processor running at 2.3 GHz and 128 GB of RAM.
Software Dependencies Yes The IBM ILOG CPLEX solver (version 12.6.1) was used to solve the proposed [SI], [MP1] and [MP2] formulations. The algorithms and the Independent Cascade infection model are implemented in Python.
Experiment Setup Yes For each network, to simulate an infection process, we chose at random a subset S of nodes and assumed them to be the misinformation sources. Afterwards, we ran the Independent Cascade diffusion process starting from them. ... We chose 2,3 and 5 as values for |S|, and simulated the process for 3 different random choices of the sources for each value of |S| (for a total of 9 simulations for each network). ... In the k-unbalanced cut solution that we look for, we always limit the size of the subset containing the sources to 5% (rounded to integer) of the number of nodes.