Estimating robustness in large social graphs


Given a large social graph, what can we say about its robustness? Broadly speaking, the property of robustness is crucial in real graphs, since it is related to the structural behavior of graphs to retain their connectivity properties after losing a portion of their edges/nodes. Can we estimate a robustness index for a graph quickly? Additionally, if the graph evolves over time, how this property changes? In this work, we are trying to answer the above questions studying the expansion properties of large social graphs. First, we present a measure that characterizes the robustness properties of a graph and also serves as global measure of the community structure (or lack thereof). We show how to compute this measure efficiently by exploiting the special spectral properties of real-world networks. We apply our method on several diverse real networks with millions of nodes, and we observe interesting properties for both static and time-evolving social graphs. As an application example, we show how to spot outliers and anomalies in graphs over time. Finally, we examine how graph generating models that mimic several properties of real-world graphs and behave in terms of robustness dynamics.

This is a preview of subscription content, access via your institution.

Fig. 1
Fig. 2
Fig. 3
Fig. 4
Fig. 5
Fig. 6
Fig. 7
Fig. 8
Fig. 9
Fig. 10
Fig. 11


  1. 1.

    This property simply means that the \(\sinh (\cdot )\) function retains the signs of the eigenvalues.

  2. 2.

  3. 3.

    Personal communication with Michael Ley and Florian Reitz from DBLP.

  4. 4.

    The bipartite graphs do not have odd length closed walks, and thus, the \(\mathrm{SC}\) is computed based on the even length closed walks. This happens replacing the \(\sinh (\cdot )\) function with the \(\cosh (\cdot )\) [15]. But then the \(\mathrm{SC}\) for the bipartite graphs cannot be efficiently approximated using similar ideas with the proposed \(\mathrm{NSC}_k\) (Sect. 4), because of the fact that the \(\cosh (\cdot )\) is an even function. However, our approach for bipartite graphs (Sect. 4, Proposition 4.1) overcomes this bottleneck and can be efficiently computed for large-scale graphs.


  1. 1.

    Akoglu L, McGlohon M, Faloutsos C (2010) OddBall: spotting anomalies in weighted graphs. In: PAKDD, pp 410–421

  2. 2.

    Albert R, Jeong H, Barabasi A-L (1999) Diameter of the world wide web. Nature 401:130–131

    Article  Google Scholar 

  3. 3.

    Albert R, Jeong H, Barabasi A-L (2000) Error and attack tolerance of complex networks. Nature 406(6794):378–382

    Article  Google Scholar 

  4. 4.

    Anagnostopoulos A, Brova G, Terzi E (2011) Peer and authority pressure in information-propagation models. In: PKDD, pp 76–91

  5. 5.

    Baeza-Yates RA, Ribeiro-Neto B (1999) Modern information retrieval. Addison-Wesley Longman, New York

    Google Scholar 

  6. 6.

    Barabási A-L, Albert R (1999) Emergence of scaling in random networks. Science 286(5439):509–512

    MathSciNet  Article  Google Scholar 

  7. 7.

    Bollobás B, Riordan O (2003) Robustness and vulnerability of scale-free random graphs. Internet Math 1(1):1–35

    MathSciNet  Article  MATH  Google Scholar 

  8. 8.

    Callaway DS, Newman MEJ, Strogatz SH, Watts DJ (2000) Network robustness and fragility: percolation on random graphs. Phys Rev Lett 80(25):5468–5471

    Article  Google Scholar 

  9. 9.

    Chakrabarti D, Faloutsos C (2012) Graph mining: laws, tools, and case studies. Synthesis lectures on data mining and knowledge discovery. Morgan and Claypool, San Rafael

    Google Scholar 

  10. 10.

    Chandola V, Banerjee A, Kumar V (2009) Anomaly detection: a survey. ACM Comput Surv 41(3):1–58

    Article  Google Scholar 

  11. 11.

    Chung FRK (1997) Spectral graph theory. CBMS, regional conference series in mathematics, no. 92. AMS

  12. 12.

    Cohen R, Havlin S (2010) Complex networks: structure, robustness and function. Cambridge University Press, Cambridge

    Google Scholar 

  13. 13.

    DBLP Bibliography Server (2006).

  14. 14.

    Erdös P, Renyí A (1960) On the evolution of random graphs. Publ Math Inst Hung Acad Sci 5:17–61

    MATH  Google Scholar 

  15. 15.

    Estrada E, Rodríguez-Velázquez JA (2005) Subgraph centrality in complex networks. Phys Rev E 71(5):056103

    MathSciNet  Article  Google Scholar 

  16. 16.

    Estrada E (2006) Spectral scaling and good expansion properties in complex networks. Europhys Lett 73(4):649–655

    MathSciNet  Article  Google Scholar 

  17. 17.

    Estrada E (2006) Network robustness to targeted attacks. The interplay of expansibility and degree distribution. Eur Phys J B 52:563–574

    Article  MATH  Google Scholar 

  18. 18.

    Faloutsos M, Faloutsos P, Faloutsos C (1999) On power–law relationships of the Internet topology. In: SIGCOMM, pp 251–262

  19. 19.

    Fortunato S (2010) Community detection in graphs. Phys Rep 486(3–5):75–174

    MathSciNet  Article  Google Scholar 

  20. 20.

    Golub GH, Van Loan CF (1996) Matrix computations, 3rd edn. Johns Hopkins University Press, Baltimore

    Google Scholar 

  21. 21.

    Hoory S, Linial N, Wigderson A (2006) Expander graphs and their applications. Bull Am Math Soc 43:439–561

    MathSciNet  Article  MATH  Google Scholar 

  22. 22.

    KDD Cup (2004).

  23. 23.

    Kumar R, Novak J, Tomkins A (2006) Structure and evolution of online social networks. In: KDD, pp 611–617

  24. 24.

    Lefevre K, Terzi E (2010) GraSS: Graph structure summarization. In: SDM, pp 454–465

  25. 25.

    Leskovec J, Lang K, Dasgupta A, Mahoney M (2009) Community structure in large networks: natural cluster sizes and the absence of large well-defined clusters. Internet Math 6(1):29–123

    MathSciNet  Article  MATH  Google Scholar 

  26. 26.

    Leskovec J, Chakrabarti D, Kleinberg J, Faloutsos C (2010) Kronecker graphs: an approach to modeling networks. J Mach Learn Res 11:985–1042

    MathSciNet  MATH  Google Scholar 

  27. 27.

    Leskovec J, Huttenlocher D, Kleinberg J (2010) Predicting positive and negative links in online social networks. In: WWW, pp 641–650

  28. 28.

    Leskovec J, Kleinberg J, Faloutsos C (2005) Graphs over time: densification laws, shrinking diameters and possible explanations. In: KDD, pp 177–187

  29. 29.

    Leskovec J, Kleinberg J, Faloutsos C (2007) Graph evolution: densification and shrinking diameters. ACM Trans Knowl Discov Data 1(1):1–41

  30. 30.

    Maiya AS, Berger-Wolf TY (2010) Expansion and search in networks. In: CIKM, pp 239–248

  31. 31.

    Malliaros FD, Megalooikonomou V, Faloutsos C (2012) Fast robustness estimation in large social graphs: communities and anomaly detection. In: SDM, pp 942–953

  32. 32.

    Malliaros FD, Vazirgiannis Michalis (2013) Clustering and community detection in directed networks: a survey. Phys Rep 533(4):95–142

    MathSciNet  Article  Google Scholar 

  33. 33.

    Maserrat H, Pei J (2010) Neighbor query friendly compression of social networks. In: KDD, pp 533–542

  34. 34.

    Mathioudakis M, Bonchi F, Castillo C, Gionis A, Ukkonen A (2011) Sparsification of influence networks. In: KDD, pp 529–537

  35. 35.

    McGlohon M, Akoglu L, Faloutsos C (2008) Weighted graphs and disconnected components: patterns and a generator. In: KDD, pp 524–532

  36. 36.

    Mihail M, Papadimitriou C, Saberi A (2011) On certain connectivity properties of the Internet topology. In: FOCS, pp 28–35

  37. 37.

    Mislove A, Marcon M, Gummadi KP, Druschel P, Bhattacharjee B (2007) Measurement and analysis of online social networks. In: IMC, pp 29–42

  38. 38.

    Mohar B (1989) Isoperimetric number of graphs. J Comb Theor B 47(3):274–291

    MathSciNet  Article  MATH  Google Scholar 

  39. 39.

    Newman MEJ (2003) The structure and function of complex networks. SIAM Rev 45:167–256

    MathSciNet  Article  MATH  Google Scholar 

  40. 40.

    Newman MEJ, Park J (2003) Why social networks are different from other types of networks. Phys Rev E 68:036122

    Article  Google Scholar 

  41. 41.

    Newman MEJ (2006) Finding community structure in networks using the eigenvector of matrices. Phys Rev E 74(3):036104

    MathSciNet  Article  Google Scholar 

  42. 42.

    Newman MEJ (2006) Modularity and community structure in networks. PNAS 103(23):8577–8582

    Article  Google Scholar 

  43. 43.

    Page L, Brin S, Motwani R, Winograd T (1999) The PageRank citation ranking: bringing order to the web. Technical Report, Stanford InfoLab

  44. 44.

    Richardson M, Agrawal R, Domingos P (2003) Trust management for the semantic web. In: ISWC, pp 351–368

  45. 45.

    Sala A, Cao L, Wilson C, Zablit R, Zheng H, Zhao BY (2010) Measurement-calibrated graph models for social network experiments. In: WWW, pp 861–870

  46. 46.

    Satuluri V, Parthasarathy S (2009) Scalable graph clustering using stochastic flows: applications to community. discovery. In: KDD, pp 737–746

  47. 47.

    Seshadhri C, Pinar A, Kolda TG (2013) An in-depth analysis of stochastic Kronecker graphs. JACM 60(2):13:1–13:32

    MathSciNet  Article  Google Scholar 

  48. 48.

    Toivonen H, Zhou F, Hartikainen A, Hinkka A (2011) Compression of weighted graphs. In: KDD, pp 965–973

  49. 49.

    Tsourakakis CE (2008) Fast counting of triangles in large real networks without counting: algorithms and laws. In: ICDM, pp 608–617

  50. 50.

    Tsourakakis CE (2011) Counting triangles in real-world networks using projections. Knowl Inf Syst 26:501–520

    Article  Google Scholar 

  51. 51.

    Viswanath B, Mislove A, Cha M, Gummadi KP (2009) On the evolution of user interaction in Facebook. In: WOSN, pp 37–42

  52. 52.

    Watts DJ, Strogatz SH (1998) Collective dynamics of ‘small-world’ networks. Nature 393(684):440–442

    Article  Google Scholar 

Download references


Fragkiskos D. Malliaros is a recipient of the Google Europe Fellowship in Graph Mining, and this research is supported in part by this Google Fellowship. Vasileios Megalooikonomou is partially supported by the ARMOR Project (FP7-ICT-2011-5.1-287720) that is co-funded by the European Commission under the Seventh Framework Programme and by the European Union (European Social Fund—ESF) and Greek national funds through the Operational Program “Education and Lifelong Learning” of the NSRF—Research Funding Program: Thales. Investing in knowledge society through the European Social Fund. Christos Faloutsos is supported by the National Science Foundation under Grants No. IIS-1217559 CNS-1314632, by the Army Research Laboratory under Cooperative Agreement Number W911NF-09-2-0053 and under Contract Number W911NF-11-C-0088, by an IBM Faculty Award and a Google Focused Research Award. Any opinions, findings and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the National Science Foundation, or other funding parties. The U.S. Government is authorized to reproduce and distribute reprints for Government purposes notwithstanding any copyright notation here on.

Author information



Corresponding author

Correspondence to Fragkiskos D. Malliaros.



In this Appendix, we provide a more detailed description of how the property of large spectral gap along with the subgraph centrality measure leads to the measure \(\xi (G)\) [17] as presented in Sect. 3. First of all, the subgraph centrality measure is defined as [15]

$$\begin{aligned} \mathrm{SC}(i) = \sum _{\ell = 0}^{\infty } \dfrac{A^\ell _{ii}}{\ell !},\quad \forall i \in V, \end{aligned}$$

where the diagonal entry \(A_{ii}\) of the matrix \(\mathbf {A}^\ell \) contains the number of closed walks of length \(\ell \) that begin and end at the same node \(i\). Focusing on unipartite graphs and keeping only the odd length closed walksFootnote 4 In order to avoid cycles in acyclic graphs, the \(\mathrm{SC}\) can be expressed as

$$\begin{aligned} \mathrm{SC}(i) = u_{i1}^2 \sinh (\lambda _1) + \sum _{j=2}^{|V|} u_{ij}^2 \sinh (\lambda _j). \end{aligned}$$

If the graph has good expansion properties (and thus high robustness), it means that \(\lambda _1 \gg \lambda _2\), and then \( u_{i1}^2 \sinh (\lambda _1) \gg \sum _{j=2}^{|V|} u_{ij}^2 \sinh (\lambda _j)\). Thus, Eq. (6) could be written as

$$\begin{aligned} \mathrm{SC}(i) \approx u_{i1}^2 \sinh (\lambda _1), ~ \forall i \in V. \end{aligned}$$

This means that for graphs with high robustness, the principal eigenvector \(u_{i1}\) will be related to \(\mathrm{SC}(i)\) as

$$\begin{aligned} u_{i1} \propto \sinh ^{-1/2}(\lambda _1) ~ \mathrm{SC}(i)^{1/2}. \end{aligned}$$

This relation suggests that if the graph shows high robustness, \(u_{i1}\) will be proportional to \(\mathrm{SC}(i)\) and a log–log plot of \(u_{i1}\) versus \(\mathrm{SC}(i), ~ \forall i \in V\) will show a linear fit with slope \(1/2\) (the discrepancy plot).

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Malliaros, F.D., Megalooikonomou, V. & Faloutsos, C. Estimating robustness in large social graphs. Knowl Inf Syst 45, 645–678 (2015).

Download citation


  • Network robustness
  • Expansion properties
  • Temporal evolution
  • Graph generating models
  • Social network analysis
  • Graph mining