#### DMCA

## A workflow for differentially-private graph synthesis (2012)

Citations: | 3 - 1 self |

### Citations

618 | Calibrating noise to sensitivity in private data analysis
- Dwork, McSherry, et al.
- 2006
(Show Context)
Citation Context ... Security, Algorithms, Measurement Keywords Differential Privacy, Graphs, Privacy, Social Networks 1. INTRODUCTION Despite recent advances in query languages [8, 12] that support differential-privacy =-=[2]-=-, several emerging areas remain underserved by these languages. Perhaps the most notable is social graph analysis, where edges in the graph reflect private information between the nodes. Informally, P... |

257 | Graph Evolution: Densification and Shrinking Diameters
- Leskovec, Kleinberg, et al.
- 2007
(Show Context)
Citation Context ...5ɛ =0.5. Due to space limitations, we only present results for two graphs; a graph of autonomous systems in the ARIN region [1] (available on our project website), the Arxiv GR-QC collaboration graph =-=[7]-=-. Statistics about the original graphs are in Table 1. In Figures 2, 1, we plot measured degree sequences, both before regression and after regression, and compare them to the degree sequence of the a... |

164 | Smooth sensitivity and sampling in private data analysis
- Nissim, Raskhodnikova, et al.
- 2007
(Show Context)
Citation Context ...original AS graph (Table 1). Therefore, we only present results for the collaboration graph, 17Assortativity 0.7 0.6 0.5 0.4 0.3 0.2 49 676 961 1600 No Buckets assortativity) [13], triangle counting =-=[9]-=-, and some generalizations of triangles [5]. We can provide analogues of each of these approaches in Weighted PINQ, typically matching proven bounds (within constants) and always exploiting the non-un... |

132 |
Privacy integrated queries: an extensible platform for privacy-preserving data analysis. InSIGMOD
- McSherry
- 2009
(Show Context)
Citation Context ...tion Services]: Data Sharing General Terms Security, Algorithms, Measurement Keywords Differential Privacy, Graphs, Privacy, Social Networks 1. INTRODUCTION Despite recent advances in query languages =-=[8, 12]-=- that support differential-privacy [2], several emerging areas remain underserved by these languages. Perhaps the most notable is social graph analysis, where edges in the graph reflect private inform... |

73 | Airavat: security and privacy for MapReduce
- Roy, Setty, et al.
(Show Context)
Citation Context ...tion Services]: Data Sharing General Terms Security, Algorithms, Measurement Keywords Differential Privacy, Graphs, Privacy, Social Networks 1. INTRODUCTION Despite recent advances in query languages =-=[8, 12]-=- that support differential-privacy [2], several emerging areas remain underserved by these languages. Perhaps the most notable is social graph analysis, where edges in the graph reflect private inform... |

51 | Relationship privacy: output perturbation for queries with joins
- Rastogi, Hay, et al.
- 2009
(Show Context)
Citation Context ...form (each edge has a “source” and “destination” attribute), graph queries typically result in many Join operations over the tables, requiring excessive amounts of additive noise using standard tools =-=[11]-=-. Bespoke analyses have recently emerged for degree distributions [3], joint degree distribution (and 7. FUTURE WORK In this short paper, we provided an overview of our new workflow for differentially... |

44 | Accurate estimation of the degree distribution of private networks
- Hay, Li, et al.
- 2009
(Show Context)
Citation Context ...); var degSeqCounts = degSeq.NoisyCount(epsilon); This query, illustrated in the figure above, is actually a Weighted PINQ implementation of an ɛ-differential privacy algorithm proposed by Hay et al. =-=[3]-=-! Query complexity. At every point in these two queries we have at most |E| items, resulting in a storage complexity of |E|. Each transformation takes linear time, so running these queries on a new (p... |

19 |
Cyclops: The Internet AS-level observatory
- Chi, Oliveira, et al.
- 2008
(Show Context)
Citation Context ...n [10]). Setting ɛ =0.1, these synthetic graphs use a total privacy cost of 5ɛ =0.5. Due to space limitations, we only present results for two graphs; a graph of autonomous systems in the ARIN region =-=[1]-=- (available on our project website), the Arxiv GR-QC collaboration graph [7]. Statistics about the original graphs are in Table 1. In Figures 2, 1, we plot measured degree sequences, both before regre... |

17 | Sharing graphs using differentially private graph models
- Sala, Zhao, et al.
(Show Context)
Citation Context ... when dmax is large, slathering on noise like this ruins the accuracy of our results. Is it really necessary to add so much noise to all entries of the JDD? Happily, the answer is no. The analysis of =-=[13]-=- shows that the noise required to protect the privacy of the JDD can be non-uniform: foreach(d1,d2) entryofthe JDD, it suffices to add noise proportional to 4 max(d1,d2). Indeed, one consequence of ou... |

16 | Probabilistic inference and differential privacy
- Williams, O, et al.
- 2010
(Show Context)
Citation Context ...ed using only the differentially private measurements taken from it. When these measurements are sufficient, we can report them and stop. However, we can go much further using probabilistic inference =-=[14]-=-, i.e., by fitting a random graph to our measurements. While our measurements are noisy, they constrain the set of plausible graphs that could lead to them. Moreover, the properties of this set of pla... |

10 | Isotone Optimization in R: Pool-Adjacent-Violators Algorithm (PAVA) and Active Set Methods
- Leeuw, Hornik, et al.
- 2009
(Show Context)
Citation Context ...quence & CCDF Hay et al. observed that a significant amount of noise can be “cleaned up” in the degree sequence by using isotonic regression (because the degree sequence is known to be nonincreasing) =-=[3, 6]-=-. We observe that the same is true for the CCDF, and moreover the degree sequence and CCDF give accurate information about different aspects of the graph: the former accurately reports the graph’s hig... |

4 |
Grigory Yaroslavtsev. Private analysis of graph structure
- Karwa, Raskhodnikova, et al.
(Show Context)
Citation Context ...only present results for the collaboration graph, 17Assortativity 0.7 0.6 0.5 0.4 0.3 0.2 49 676 961 1600 No Buckets assortativity) [13], triangle counting [9], and some generalizations of triangles =-=[5]-=-. We can provide analogues of each of these approaches in Weighted PINQ, typically matching proven bounds (within constants) and always exploiting the non-uniformities described in Section 1.1. Many g... |

1 |
Evimaria Terzi. Tutorial on privacy-aware data management in information networks
- Hay, Liu, et al.
- 2011
(Show Context)
Citation Context ...ypically matching proven bounds (within constants) and always exploiting the non-uniformities described in Section 1.1. Many graph analyses satisfy privacy definitions other than differential privacy =-=[4]-=-. These definitions generally do not exhibit the robustness of differential privacy, and a comparison is beyond the scope of this note. 0.1 0 0 0.5 1 1.5 2 MCMC Steps x 10 6 Figure 4: Assort’y vs MCMC... |