For more information about this meeting, contact Jason Morton, Carina Curto, Vladimir Itskov.
|Title:||Inference using noisy degrees: Differentially Private beta-model and synthetic graphs|
|Seminar:||Applied Algebra and Network Theory Seminar|
|Speaker:||Sesa Slavkovic, Penn State|
|The β-model of random graphs is an exponential family model with the degree sequence as a sufficient statistic. Motivated by data privacy problems with network data, we present a differentially private estimator of the parameters of β-model. We show that the estimator is asymptotically consistent and achieves the same rate as the non private estimator. Our techniques involve releasing the degree sequence using Laplace mechanism and constructing a maximum likelihood estimate of the degree sequence, which is equivalent to projecting the noisy degree sequence on the set of all graphical degree sequences. We present an efficient algorithm for the projection which also outputs a synthetic graph. Our techniques can also be used to release degree distributions accurately and privately, and to estimate noisy degrees arising from contexts other than the privacy. We evaluate the proposed estimator on real graphs and compare it with a current algorithm for releasing degree distributions and find that it does significantly better.|
Room Reservation Information
|Date:||11 / 05 / 2014|
|Time:||02:30pm - 03:20pm|