LBS based framework to obstruct linking attack in data releases

Big data repository is a myriad collection of data that needs enormous processing for data analysis to extract interesting patterns. It involves third party to do the data analysis that endangers the privacy of the data. Hence, along with efficient real time analysis of data, privacy of the data also needs a careful attention. In this paper, we discuss the problem of linking attack in the data releases. We show the attack due to the linking of social media data with relational sensitive database discloses the sensitive information about a social entity. Though the data releases don’t contain Identifiers (IDs) information like Name, Social Security Number (SSN), etc., linking attack is still possible through the combination of background knowledge with Quasi-Identifiers (QIDs) information like age, sex, etc. The solution encourages the release of anonymized relational sensitive database that obstruct the linking with other databases like social media data. Anonymization is a way to transform the original data to an anonymized version, such that the linkage between the sensitive information of an entity is dissociated from his/her identity. Most of the anonymity models are based on the concept of k-anonymity and l-diversity. In this paper, we present an extensive study on Greedy based heuristics. We propose Local Beam Search (LBS) based global generalization approach to achieve k-anonymity with l-diversity. The experimental evaluation confirms that the proposed method outperforms the existing Greedy based heuristics. The proposed method guarantees a faster convergence than existing Greedy based heuristics. Also, the proposed method performs better than the existing (t, k) Hypergraph Anonymization.