As i have read many literatures/papers on Privacy in big data, i got to know that various people are using different technique to achieve it. They are:
1. Anonymization Technique
2. Differential Privacy
3. Homomorphic Encryption
All of above have its own pros and cons, so kindly suggest me which is best to choose for providing privacy in Big Data Analytics?
Also where to provide privacy??? At data generation phase, data collection Phase or data analytics phase???
Differential Privacy is better compared to remaining method as per my knowledge.
Each of those methodologies has its appropriate uses/situations. DP is primarily for statistical queries whereas secure computation (e.g., HE, Garbled Circuits, etc.) will have a wider range of application. I would consider the application of all of these technologies in concert.
For instance, if you look at the TIPPERS program at UCI (http://tippersweb.ics.uci.edu/) you will find an excellent example of differential privacy and other secure computation technologies being used together.
The Homomorphic Encryption Standardization Consortium (http://homomorphicencryption.org/) has a good report on applications and use cases that may be of interest, as well.
Interesting work you may not be aware of: Kung, S.Y., Chanyaswad, T., Chang, J.M. and Wu, P., 2017. Collaborative PCA/DCA learning methods for compressive privacy. ACM Transactions on Embedded Computing Systems (TECS), 16(3), p.76.
I am running a workshop this summer on big data privacy and security and I hope you'll consider submitting a paper: https://www.ares-conference.eu/workshops/spebd-2018/
- Badges
- Science topic