Association Rule Hiding for Privacy Preserving Data Mining: A Survey on Algorithmic Classifications

摘要

The increased collection, storage and analysis of person-specific data cause serious challenges to the protection of the identities to which such data correspond. In various conditions, the extracted knowledge is highly confidential and it needs to be sanitized before published in order to address privacy concerns. Data mining technology is capable of extract huge amount of knowledge with minimal time period. The knowledge extracted by intelligent data mining algorithm may reveal most sensitive information belongs to a person or an organization. Data belongs to a person or an organization may have different sensitive levels. These data are made available only for authorized persons. So ensuring the protection of sensitive data by access restriction is not a complete method. This may affect the utility of the data mining result and with help of the knowledge the user may reidentify sensitive data items from non-sensitive data is known as 'Inference Problem'. The privacy preserving data mining is to provide a solution for protecting sensitive information by developing a data mining techniques which could be applied on databases without affecting the accuracy of data mining result. At the same time without violating the privacy of individuals. This paper states a detailed study on various algorithms for association rule hiding methods.