Diffusion Wavelet-based Privacy Preserving in Social Networks
Author(s)
Tian, Hui
Liu, Jingtian
Shen, Hong
Griffith University Author(s)
Year published
2018
Metadata
Show full item recordAbstract
Data in social networking services (SNS) are spreading widely at high speed and always have significant impacts. It is very vulnerable to attacks. Many segments of information can be defined as user’s privacy. In a friends’ network, subgraph structure of nodes could be the basis for an adversary identifying the target node. In a computer network, a link’s weight between two nodes represents how often they communicate between each other. Attackers would collect these personal data and analyze them by various data mining methods. There are mainly two categories of targets, node re-identification, and linkage information. In ...
View more >Data in social networking services (SNS) are spreading widely at high speed and always have significant impacts. It is very vulnerable to attacks. Many segments of information can be defined as user’s privacy. In a friends’ network, subgraph structure of nodes could be the basis for an adversary identifying the target node. In a computer network, a link’s weight between two nodes represents how often they communicate between each other. Attackers would collect these personal data and analyze them by various data mining methods. There are mainly two categories of targets, node re-identification, and linkage information. In order to protect different kinds of sensitive private information, it is demanding to establish appropriate privacy preserving models and develop efficient methods to protect private data before publishing. In many existing models, node degree and subgraph information are the sensitive information which may be used by adversary parties in node/linkage disclosure. They assumed the adversary has partial information from open queries. Different techniques have been applied in privacy preserving before publishing graph data, such as dynamic programming and forming super-node to guarantee k-anonymity and l-diversity [23,24] and differential privacy based techniques [1,12,13]. However, these methods are either identified as NP-hard problems or convergent to locally optimized solution. The data utility is low after publishing.
View less >
View more >Data in social networking services (SNS) are spreading widely at high speed and always have significant impacts. It is very vulnerable to attacks. Many segments of information can be defined as user’s privacy. In a friends’ network, subgraph structure of nodes could be the basis for an adversary identifying the target node. In a computer network, a link’s weight between two nodes represents how often they communicate between each other. Attackers would collect these personal data and analyze them by various data mining methods. There are mainly two categories of targets, node re-identification, and linkage information. In order to protect different kinds of sensitive private information, it is demanding to establish appropriate privacy preserving models and develop efficient methods to protect private data before publishing. In many existing models, node degree and subgraph information are the sensitive information which may be used by adversary parties in node/linkage disclosure. They assumed the adversary has partial information from open queries. Different techniques have been applied in privacy preserving before publishing graph data, such as dynamic programming and forming super-node to guarantee k-anonymity and l-diversity [23,24] and differential privacy based techniques [1,12,13]. However, these methods are either identified as NP-hard problems or convergent to locally optimized solution. The data utility is low after publishing.
View less >
Journal Title
Computers and Electrical Engineering
Volume
67
Subject
Software engineering