SENSITIVE PATTERN HIDING IN DYNAMIC DATASETS

Abstract

Data Mining deals with extraction of previously unknown patterns from large amounts of data sets. In today’s data driven world, every interaction results in huge amounts of data being generated. With the help of this data businesses have developed a more customer centric approach over the years. The business organizations perform analytics on the data generated by them collaboratively to find existing trends and patterns in the transactional data which can help them further to enhance their strategies. The analysis being performed on the data may lead to the disclosure of some patterns/ trends which are sensitive to the organization or the individual. This may include sensitive individual information or critical business information. In order to minimise or completely avoid the disclosure of any such sensitive information, privacy preserving data mining is used. Most of the existing techniques for hiding sensitive patterns or trends in transactional data work on static data, i.e. the entire data is present before the processing begins and there is no change in data. The data that is being generated today is dynamic in nature. New data is continuously added and the old data may become irrelevant. Frequent patterns are generated in the dynamic dataset with the help of FP Growth algorithm. The FP Tree is generated and is continuously updated to get the frequent itemsets in the current time slot.