Releasing evolving networks which contain sensitive information could compromise individual privacy. In this paper, we study the problem of releasing evolving networks under differential privacy. We explore the possibility of designing a differentially private evolving networks releasing algorithm. We found that the majority of traditional methods provide a snapshot of the networks under differential privacy over a brief period of time. As the network structure only changes in local part, the amount of required noise entirely is large and it leads to an inefficient utility. To this end, we propose GHRG-DP, a novel differentially private evolving networks releasing algorithm which reduces the noise scale and achieves high data utility. In the GHRG-DP algorithm, we learn the online connection probabilities between vertices in the evolving networks by generalized hierarchical random graph (GHRG) model. To fit the dynamic environment, a dendrogram structure adjusting method in local areas is proposed to reduce the noise scale in the whole period of time. Moreover, to avoid the unhelpful outcome of the connection probabilities, a Bayesian noisy probabilities calculating method is proposed. Through formal privacy analysis, we show that the GHRG-DP algorithm is -differentially private. Experiments on real evolving network datasets illustrate that GHRG-DP algorithm can privately release evolving networks with high accuracy.

中文翻译：

---

基于GHRG的演进网络的差分隐私

释放包含敏感信息的不断发展的网络可能会损害个人隐私。在本文中，我们研究了在差分隐私下释放不断发展的网络的问题。我们探索了设计差分私有演进网络发布算法的可能性。我们发现大多数传统方法都可以在很短的时间内提供不同隐私下的网络快照。由于网络结构仅在局部发生变化，因此所需的噪声总量很大，从而导致效用低下。为此，我们提出了GHRG-DP，这是一种新颖的差分专用演进网络发布算法，可减少噪声规模并实现高数据实用性。在GHRG-DP算法中，广义分层随机图（GHRG）模型。为了适应动态环境，提出了一种局部树状图结构调整方法，以减小整个时间段的噪声规模。此外，为了避免连接概率的不利结果，提出了一种贝叶斯噪声概率计算方法。通过正式的隐私分析，我们表明GHRG-DP算法是-差分私有的。在真实的演进网络数据集上进行的实验表明，GHRG-DP算法可以高精度地私下发布演进的网络。