Providing anonymity using top down specialization on Big Data using hadoop framework

摘要

Sharing of data has become a new trend among internet users which has lead to issue of privacy. Hence to provide privacy to user's data anonymity is provided. Data anonymity is a process of hiding sensitive information which is responsible for breach of privacy. The number of users sharing data has increased tremendously which has to lead to generation of huge data. This huge data that cannot be managed by normal system and software is termed as ???Big data???. Existing system and anonymity approaches fail to handle Big Data efficiently. Big data handling is complex issue as to perform any operation system must be capable of manipulating such huge data in acceptable time. System must be highly scalable. This paper provides anonymity for Big Data in a highly scalable fashion. It makes use of MapReduce framework which gains scalability by job level and task level parallelization. K-anonymity is used to provide anonymity which generalises the data. Job level parallelization refers to running of multiple MapReduce jobs simultaneously. Task level parallelization refers to running of multiple mapper/reducer over data splits. To make full use of parallel computation anonymization process is split into two phases. In first phase the huge data set is portioned into small data sets and anonymity is provided. But the data obtained is inconsistent hence second phase is executed which merges this anonymous data into one single huge data set. This paper accomplishes specialization computation in highly scalable manner.