Author : Johny Antony P 1
Date of Publication :7th October 2016
Abstract: We are in the midst of big data. The rate of data generation is increasing at a very rapid rate. We need to understand and analyze this data as quick as possible. A delay in millisecond to understand the data may cost not only money but also life. There are various processing and analytic mechanisms like Hadoop and MapReduce to process the data. But as big data comprises an enormous amount of personally identifiable information, user privacy and security is a major concern, and it is a massive challenge in big data. It is considered as an absolute prerequisite for exchanging sensitive information in terms of analysis, validation and publishing. The multidimensional anonymization and access control are widely-adopted privacy preservation approaches. Despite much research a method with satisfactory privacy settings are far from being achieved. Owing to the lack of integrating data from multiple sources, manual administration, video surveillance applications, the traditional methods are not feasible to big data. Hence, scalability is concerned as the major problem encountered when the conventional preservation techniques are applied to the big data. Some of the approaches have handled the security and privacy problems at the time of data shared among the different organizations. However, they do not efficiently preserve the data privacy since they fail in handling the attacks. In this paper we present a new frame work for preserving privacy using the basic concepts of differential privacy and overlapped slicing.
Reference :