Review
Minimizing the spread of misinformation in online social networks: A survey

https://doi.org/10.1016/j.jnca.2021.103094Get rights and content

Abstract

Online social networks provide an opportunity to spread messages and news fast and widely. One may appreciate the quick spread of legitimate news and messages but misinformation can also be spread quickly and may raise concerns, questioning reliability and trust in such networks. As a result, detecting misinformation and containing its spread has become a hot topic in social network analysis. When misinformation is detected, some actions may be necessary to reduce its propagation and impact on the network. Such actions aim to minimize the number of users influenced by misinformation. This paper reviews approaches for solving this problem of minimizing spread of misinformation in social networks and proposes a taxonomy of different methods.

Introduction

The proliferation of internet technologies has led to an increasing number of online social networks and users. Several indicators suggest that the number of users keeps increasing and a large number of people have accepted online social networks as a major source of news. The potential of social networks has led to significant research in trying to propagate news widely by identifying so-called influential users (Zareie et al., 2019, Sheikhahmadi and Zareie, 2020). This problem, known as Influence Maximization (Domingos and Richardson, 2001, Kempe et al., 2003), has attracted lots of attention recently. However, spreading news fast gives rise to an adverse effect: misinformation can be spread fast too.

Users may spread misinformation inadvertently or with different financial and social motivations (Meel and Vishwakarma, 2019). Misinformation propagation has become a significant threat in social networks and reduces the reliability and confidence of the users towards news and messages. As quoted in Tong et al. (2018b), a report published by the World Economic Forum regards spreading misinformation as one of the top global economic risks. Spreading misinformation or unsubstantiated rumours may have widely negative impact and may lead to economic damages, significant disruption or even widespread panic; various examples have been mentioned in the literature (Wen et al., 2015, Fan et al., 2013, Wu et al., 2017, Fan et al., 2014, Tong et al., 2020, Fang et al., 2020). Misinformation may take different forms. In this paper, the term misinformation is used as a general term to refer to any false or inaccurate information which may be spread in online social networks intentionally or unintentionally.

Confronting misinformation in social networks has attracted lots of attention among researchers. To deal with it, there are several aspects that have to be addressed: (i) identification of misinformation among newly produced information, also known as misinformation detection, is important as early detection of misinformation decreases the chances of wide propagation with potentially adverse effects; (ii) detection of the sources of misinformation is important as it helps identify the culprits and potentially malicious users who initiate the propagation of misinformation; (iii) tracking subsequent (re)appearances of already detected misinformation, possibly in a slightly altered form but still misinformation; (iv) minimization of the spread of misinformation is another important aspect which aims to prevent the propagation of misinformation in the network. A plethora of methods have been proposed for each of these aspects in the literature.

Methods to detect misinformation and identify sources are reviewed in Bondielli and Marcelloni, 2019, Meel and Vishwakarma, 2019, Alzanin and Azmi, 2018 and Meel and Vishwakarma, 2019, Shelke and Attar, 2019, respectively. Methods to detect misinformation are based on mapping a stream of social media posts to a classification system that labels posts as misinformation or non-misinformation. Methods to identify sources are based on the network structure and propagation graph from which users or locations that initiate misinformation are identified. In Sharma et al. (2019), intervention methods for misinformation detection and mitigation are classified and reviewed. Approaches for the development of data mining tools for misinformation tracking and verification are reviewed in Zubiaga et al. (2018). In Ahsan et al. (2019), methods for detection and controlling rumour in social networks are reviewed from a multidisciplinary (Psychology, Sociology and Epidemiology) viewpoint; Ahsan et al. (2019) also reviews the features that favour wide propagation of misinformation. Yet, the literature lacks a comprehensive review and classification of the methods explicitly proposed to minimize the spread of misinformation, which act as an important deterrent when confronting misinformation. Although (Meel and Vishwakarma, 2019, Ahsan et al., 2019, Sharma et al., 2019) shortly pay attention to this topic, they do not focus on minimizing the spread of misinformation. This paper aims to fill this gap by reviewing and classifying all existing methods in the literature for the minimization of the spread of misinformation. In comparison with previous surveys, distinct differences of our work are:

  • We focus on approaches that minimize the spread of misinformation in social networks, after misinformation has been detected.

  • A new taxonomy and a comprehensive review of state-of-the-art methods is presented that offers extensive coverage of the subject.

  • Evaluation strategies that include real-world datasets and random models to generate synthetic datasets for evaluation purposes are also presented.

  • Current challenges and potential future directions are thoroughly discussed.

The rest of the paper is organized as follows: Section 2 contains definitions and background information. A formal definition of the problem of Minimizing the Spread of Misinformation (MSM) and key strategies to address it are presented in Section 3. A detailed discussion of the different methods to find solutions to the MSM problem is given in Sections 4 Blocking-based methods, 5 Clarification-based methods . Section 6 covers evaluation strategies and datasets for the assessment of different methods. Finally, Section 7 concludes the paper and discusses future research directions.

Section snippets

Online social networks

An online social network is an abstraction that captures the interactions between people relying on some internet-based infrastructure. People join online social networks with different goals, such as socializing, keeping in touch with friends, as well as reading and/or sharing news. The ability of every user to spread news is an important benefit of online social networks but it has an adverse effect too. Alongside legitimate information, spreading misinformation may have some disruptive

The problem of minimizing the spread of misinformation

Different approaches have been utilized to detect misinformation  (Bondielli and Marcelloni, 2019). Independent of these approaches, once misinformation is detected, a containment strategy should be adopted to minimize the spread of misinformation. In brief, the problem of Minimizing the Spread of Misinformation (MSM) can be defined as follows. A set of Malicious Nodes (MN) intends to propagate misinformation in a social network. A solution to the MSM problem aims to minimize the number of

Blocking-based methods

This section covers methods relying on blocking nodes or edges. As discussed, in a blocking strategy a set of nodes or edges are removed to minimize the spread of misinformation in the network.

Clarification-based methods

In these methods, once again, the assumption is that misinformation originates from a set of certain malicious nodes, MN. However, the aim is to identify a set of nodes, TN, to initiate a truth campaign, that is, to spread a clarification message that will counter the misinformation originating from MN. The ultimate goal is to minimize the number of users accepting (or influenced by) the misinformation. It is noted that this problem, first modelled by He et al. in He et al. (2012), is different

Evaluation strategies and datasets

All methods proposed to solve the problem need to be evaluated regarding their performance. In principle, evaluation aims to assess the impact of a method on minimizing the number of nodes of a graph that will be influenced by misinformation. Different graphs are used, which include both synthetic graphs and graphs based on datasets from real-world networks.

In order to evaluate node or edge blocking methods, some nodes are randomly determined as malicious nodes and their spreading ability (that

Conclusion and future directions

In this paper, proposed methods for misinformation spread minimization were classified into two categories. In blocking-based methods the idea is to change the network structure; some nodes and/or edges are removed from the network to minimize the spread of misinformation. Blocking nodes and/or edges comes with a cost and may lead to a discredit of the network if it takes a long time to restore connectivity of nodes and edges. In clarification-based methods, the goal is to increase the

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Ahmad Zareie is currently studying towards a Ph.D. in the Information Management Group of Department of Computer Science at the University of Manchester. He is generally interested in data mining, graph analysis, and machine learning. His current work focus on networks, including social networks analysis, complex networks and distributed systems.

References (105)

  • ShelkeS. et al.

    Source detection of rumor in social network–a review

    Online Soc. Netw. Media

    (2019)
  • WuQ. et al.

    Influence of dynamic immunization on epidemic spreading in networks

    Physica A

    (2015)
  • WuP. et al.

    Scalable influence blocking maximization in social networks under competitive independent cascade models

    Comput. Netw.

    (2017)
  • WuQ. et al.

    Rumor restraining based on propagation prediction with limited observations in large-scale social networks

  • YangL. et al.

    Containment of rumor spread in complex social networks

    Inform. Sci.

    (2020)
  • YangD. et al.

    Dynamic node immunization for restraint of harmful information diffusion in social networks

    Physica A

    (2018)
  • ZareieA. et al.

    Identification of influential users in social networks based on users’ interest

    Inform. Sci.

    (2019)
  • ArazkhaniN. et al.

    Influence blocking maximization in social network using centrality measures

  • BarabásiA.-L. et al.

    Emergence of scaling in random networks

    Science

    (1999)
  • BharathiS. et al.

    Competitive influence maximization in social networks

  • BorgsC. et al.

    Maximizing social influence in nearly optimal time

  • BorodinA. et al.

    Threshold models for competitive influence in social networks

  • BudakC. et al.

    Limiting the spread of misinformation in social networks

  • BuscarinoA. et al.

    Disease spreading in populations of moving agents

    Europhys. Lett.

    (2008)
  • CarnesT. et al.

    Maximizing influence in a competitive social network: a follower’s perspective

  • ChelkakD. et al.

    Universality in the 2D Ising model and conformal invariance of fermionic observables

    Invent. Math.

    (2012)
  • ChenT. et al.

    Minimizing misinformation profit in social networks

    IEEE Trans. Comput. Soc. Syst.

    (2019)
  • ChenW. et al.

    Scalable influence maximization for prevalent viral marketing in large-scale social networks

  • De DomenicoM. et al.

    The anatomy of a scientific rumor

    Sci. Rep.

    (2013)
  • DeyP. et al.

    Centrality based information blocking and influence minimization in online social network

  • DomingosP. et al.

    Mining the network value of customers

  • ErdosP. et al.

    On the evolution of random graphs

    Publ. Math. Inst. Hung. Acad. Sci.

    (1960)
  • FanL. et al.

    Least cost rumor blocking in social networks

  • FanL. et al.

    Maximizing rumor containment in social networks with constrained time

    Soc. Netw. Anal. Min.

    (2014)
  • GoldenbergJ. et al.

    Using complex systems analysis to advance marketing theory development: Modeling heterogeneity effects on new product growth through stochastic cellular automata

    Acad. Mark. Sci. Rev.

    (2001)
  • GranovetterM.

    Threshold models of collective behavior

    Am. J. Sociol.

    (1978)
  • HeX. et al.

    Influence blocking maximization in social networks under the competitive linear threshold model

  • HolmeP. et al.

    Attack vulnerability of complex networks

    Phys. Rev. E

    (2002)
  • HosniA.I.E. et al.

    DARIM: Dynamic approach for rumor influence minimization in online social networks

  • HosniA.I.E. et al.

    Hisbmodel: a rumor diffusion model based on human individual and social behaviors in online social networks

  • HosniA.I.E. et al.

    Least cost rumor influence minimization in multiplex social networks

  • KempeD. et al.

    Maximizing the spread of influence through a social network

  • Khalil, E., Dilkina, B., Song, L., 2013. CuttingEdge: influence minimization in networks. In: Proceedings of the...
  • KhalilE.B. et al.

    Scalable diffusion-aware optimization of network topology

  • KimuraM. et al.

    Minimizing the spread of contamination by blocking links in a network

  • KimuraM. et al.

    Solving the contamination minimization problem on networks for the linear threshold model

  • KimuraM. et al.

    Blocking links to minimize contamination spread in a social network

    ACM Trans. Knowl. Discov. Data

    (2009)
  • KimuraM. et al.

    Extracting influential nodes for information diffusion on a social network

  • KuhlmanC.J. et al.

    Blocking simple and complex contagion by edge removal

  • LargeronC. et al.

    Dancer: dynamic attributed networks with community structure generation

    Knowl. Inf. Syst.

    (2017)
  • Cited by (0)

    Ahmad Zareie is currently studying towards a Ph.D. in the Information Management Group of Department of Computer Science at the University of Manchester. He is generally interested in data mining, graph analysis, and machine learning. His current work focus on networks, including social networks analysis, complex networks and distributed systems.

    Rizos Sakellariou obtained his Ph.D. in Computer Science from the Victoria University of Manchester in 1997. Following positions with Rice University and the University of Cyprus he has been an academic member of staff of the University of Manchester since 2000 where he is currently Professor of Computer Science. He has published over 150 research papers on a range of topics primarily related to distributed systems, his research has been supported by more than 20 UK and EU projects and has served on the Program Committee of over 160 conferences and workshops. He values integrity, collaboration and a strong work ethic.

    View full text