Geo-social network publication based on differential privacy
Geo-social network publication based on differential privacy
复制标题
基于差分隐私的地理社交网络发布
DOI：
10.1007/s11704-018-8075-z
复制
发表时间：
2018-11
期刊：
影响因子：
--
通讯作者：
Yidong Li
中科院分区：
文献类型：
--
作者：
Xiaochun Wang;Yidong Li
随着移动的技术的发展，地理社会网络（Geo-social network，GSN）数据被频繁地收集和发布以供数据收集者进一步分析，从而提供个性化服务，而发布的数据集可能包含隐私和敏感信息。即使一些数据收集者应用了一些简单的基于匿名化的技术，由于对手从已发布的GSN获得的各种异构背景知识，用户和位置的敏感信息仍然可能以很高的可能性被公开。因此，隐私保护是GSN发布的瓶颈。众所周知，从真实的世界中采集的数据往往具有多方关系，并以超图的形式表示。而传统的图-差分隐私算法（DP）由于不能捕获异构背景知识攻击，且对超图表示的数据中的高阶关系提供隐私的能力不足，不适合为超图提供隐私。最近，一些基于差分隐私[1，2]和加密机制[3 - 5]的研究试图为GSN数据发布提供隐私保护，但在减少噪声注入方面面临挑战。受Li等人[6]工作的启发，本文首先提出了一种用于超图表示的GSN发布的差分隐私模型，该模型能够为GSN中的多方关系提供强隐私。我们将差分隐私的概念从图数据扩展到超图数据，实现了位置和社交两方面的差分隐私
With the development of mobile technology, geo-social network (GSN) data is frequently collected and released for further analysis by data collectors in order to provide personalized service, while the released dataset may contain private and sensitive information. Even though some data collectors apply a number of simple anonymization-based techniques, sensitive information of users and locations may be still disclosed with high possibility because of the variety of heterogeneous background knowledge an adversary obtains from the published GSN. Therefore, privacy preserving is a bottleneck in GSN publishing. It is well known that data collected from real world often has multiparty relationships and be presented by hypergraph. While the traditional graph-differential privacy (DP) is not suitable for providing privacy to hypergraph because it can not capture the heterogeneous background knowledge attack and is deficient to provide privacy for high-order relationship in hypergraph-represented data. Recently, some studies dased on differential privacy [1,2] and encryption mechanism [3–5] attempted to provide privacy preserving to GSN data publication, yet facing challenges in reducing noise injection. In this paper, as inspired by the works of Li et al. [6], we first propose a differential privacy model for hypergraph-represented GSN publication, which is able to provide strong privacy for the multi-party relationship in GSN. We develop the notion of differential privacy from graph data to hypergraph data and achieve differential privacy in both location and social
登录
查看更多内容
DOI：
10.1109/anti-cybercrime.2017.7905256
发表时间：
2017-03
期刊：
2017 2nd International Conference on Anti-Cyber Crimes (ICACC)
影响因子：
--
作者：
Abdullah Albelaihy;Jonathan M. Cazalas
通讯作者：
Abdullah Albelaihy;Jonathan M. Cazalas
L-EncDB：云计算中隐私保护数据查询的轻量级框架
影响因子：
8.8
作者：
Li, Jin;Liu, Zheli;Wong, Duncan S.
通讯作者：
Wong, Duncan S.
影响因子：
2.7
作者：
P. Xiong;Tianqing Zhu;Wenjia Niu;Gang Li
通讯作者：
P. Xiong;Tianqing Zhu;Wenjia Niu;Gang Li
影响因子：
8.1
作者：
Gao, Chong-zhi;Cheng, Qiong;Li, Jin
通讯作者：
Li, Jin
保护私有地理社交网络免受异构信息的实际混合攻击
影响因子：
6
作者：
Yuechuan Li;Yidong Li
通讯作者：
Yidong Li