Recent years have seen a remarkable development of privacy-preserving computing (PPC) as an effective way to achieve secure distributed computing and information sharing across a public network that are of critical importance for network-centric computation and big data analytics. Recent advancement in cloud data sharing techniques has made traditional PPC techniques unable to effectively against emerging malicious attacks such as shilling, collusion and inference attacks.
In the first part of this talk, I will first introduce the problem of privacy-preserving computing, its research challenges in cloud big data computing, then give a taxonomy on data protection techniques categorized on the security levels of data publishing, with the focus on differential privacy as an effective method to combat inference attacks, and provide an overview on our contributions in privacy-preserving computing. In the second part, to show the power of differential privacy for secure data sharing, I will give two examples of our work of applying differential privacy to achieve privacy-preserving recommendation and data clustering against inference attacks. As concluding remarks, I will further illustrate the application of differential privacy in obtaining privacy-preserving solutions for some statistical and combinatorial optimization problems.