Spectrum-Centric Differential Privacy for Hypergraph Spectral Clustering
In real world, most of complex networks can be represented by hypergraphs. Hypergraph spectral clustering has drawn wide attention in recent years due to it is more suitable to describe high-order information between objects. In this paper, we focus on the spectrum of hypergraph-based complex networks, propose a spectrum-centric differential privacy scheme by using stochastic matrix perturbation. The main idea is to project the hypergraph Laplacian matrix into a low dimensional space and perturb the eigenvector with random noise. We present a differential privacy mechanism for hypergraph clustering based on the exponential mechanism. We evaluated the computational efficiency and data utility of the existing methods on both synthetic datasets and real datasets, and the experimental results reveal that our proposed mechanism achieves a better performance in both data utility and efficiency.
KeywordsDifferential privacy Information security Hypergraph Spectral clustering
This work is supported by National Science Foundation of China Grant #61672088, Fundamental Research Funds for the Central Universities #2016JBM022 and #2015ZBJ007.
- 1.Narayanan, A., Shmatikov, V.: De-anonymizing social networks. In: 2009 30th IEEE Symposium on Security and Privacy, pp. 173–187. IEEE (2009)Google Scholar
- 3.Pihur, V., Korolova, A.: RAPPOR: randomized aggregatable privacy-preserving ordinal response. In: ACM SIGSAC Conference on Computer and Communications Security, pp. 1054–1067. ACM (2014)Google Scholar
- 4.Li, Y., Wu, X., Lu, A.: Analysis of spectral space properties of directed graphs using matrix perturbation theory with application in graph partition, pp. 847–852 (2015)Google Scholar
- 5.Kasiviswanathan, S.P., Nissim, K., Raskhodnikova, S.: Analyzing graphs with node differential privacy 7785, 457–476 (2013)Google Scholar
- 6.Day, W.-Y., Li, N., Lyu, M.: Publishing graph degree distribution with node differential privacy. In: Proceedings of the 2016 International Conference on Management of Data, SIGMOD 2016, pp. 123–138. ACM, New York (2016)Google Scholar
- 9.Seary, A.J., Richards, W.D.: Spectral methods for analyzing and visualizing networks: an introduction. na (2003)Google Scholar
- 12.Chaudhuri, K., Sarwate, A., Sinha, K.: Near-optimal differentially private principal components. In: Advances in Neural Information Processing Systems, pp. 989–997 (2012)Google Scholar
- 13.Kapralov, M., Talwar, K.: On differentially private low rank approximation. In: Proceedings of the Twenty-Fourth Annual ACM-SIAM Symposium on Discrete Algorithms, pp. 1395–1414. SIAM (2013)Google Scholar
- 14.Ahmed, F., Jin, R., Liu, A.X.: A random matrix approach to differential privacy and structure preserved social network graph publishing. arXiv preprint arXiv:1307.0475 (2013)
- 15.Wang, T., Blocki, J., Li, N., et al.: Locally differentially private protocols for frequency estimation. In: Proceedings of the 26th USENIX Security Symposium, pp. 729–745 (2017)Google Scholar
- 16.Mcsherry, F., Talwar, K.: Mechanism design via differential privacy. In: 2007 IEEE Symposium on Foundations of Computer Science, FOCS 2007, pp. 94–103 (2007)Google Scholar
- 18.Hoffman, A.J., Wielandt, H.W.: The variation of the spectrum of a normal matrix. In: Selected Papers of Alan J Hoffman: With Commentary, pp. 118–120. World Scientific (2003)Google Scholar