Privacy Preservation in IoT: Machine Learning Approaches by Youyang Qu & Longxiang Gao & Shui Yu & Yong Xiang

Author:Youyang Qu & Longxiang Gao & Shui Yu & Yong Xiang
Language: eng
Format: epub
ISBN: 9789811917974
Publisher: Springer Nature Singapore

Youyang Qu

Email: [email protected]

The cyber-physical social system (CPSS), as an extension of Internet of Things (IoT), maps human social interaction from cyberspace to the physical world by data sharing and posting, such as publishing spatial-temporal data, images, videos, etc. The published data contains individual’s sensitive information, and thereby leads to continues attacks on it. Nowadays, most existing research assumes that the privacy protection should be uniform that all parties share a same privacy protection level. This impractical assumption results in data degradation due to over-protection or privacy leakage due to under-protection. Therefore, we carry out a personalized privacy protection model by using generative adversarial nets (GAN) to enhance differential privacy (P-GAN), especially the protection on sensitive spatial-temporal data. In GAN-DP, we add a Differential Privacy Identifier to the classic GAN, in which there are only a Generator and a Discriminator. P-GAN is able to generate synthetic data can can best mimic the raw data while ensuring a satisfying privacy protection level. It can be told that P-GAN can optimize the trade-off between personalized privacy protection and improved data utility. At the same time, the learning process can achieve fast convergence. Moreover, the GAN-driven differentially private noise generation process decouples the correlation of the injected noise. In this way, collusion attacks, which is a leading attack in privacy-preserving domain, can be eliminated. We evaluate with experiments on real-world datasets, which show the upgradation of optimized trade-off and improved efficiency of P-GAN over existing research. The theories and experiments testify the effectiveness of P-GAN in IoT scenarios. This chapter is mainly based on our recent research on GAN-enhanced differential privacy [1–7].

Download

Copyright Disclaimer:
This site does not store any files on its server. We only index and link to content provided by other sites. Please contact the content providers to delete copyright contents if any and email us, we'll remove relevant links or contents immediately.