Deriving Attribute Relationship for Big Data Security

Abstract: There has been an ever rising interest towards Big Data and its security. Big Data is vast to be analysed and providing security to entire Big Data is an uphill task. Finding a solution to make sure that the Big Data is secured could be easier when we try to select only important information from the Big Data and provide security to that information only. Hence, the main goal of this paper is to propose an attribute selection methodology based on the relationship and the sensitivity among attributes for Big Data security. To be more precise, this paper basically includes two things: attribute selection methodology and provision of security. When it comes to providing security, one must understand that not all the data in Big Data needs to be secured because not every information in the Big Data is rather important. So this paper tries to select an attribute based on its relevance and the confidentiality level and provide security to that particular attribute only. That is, an attribute with higher relevance is more important than other attributes to provide security. This process includes two Datasets: User-defined Dataset and Reference Dataset. Attributes from both the Datasets are compared for their relevance and these attributes are classified into three categories namely Restricted, Confidential and Public. This becomes easier to provide security to each of these categories accordingly.

Keywords: Big Data, Machine Learning, Artificial Neural Network, Data Mining, Datasets, Security