Preserving Individual Privacy from Inference Attack in Transaction Data Publishing

Recently, industries such as retail and e-commerce companies have collaborated in analyzing transaction databases. It is suggested that the activity can bring various benefits for the companies since hidden information can be revealed from the analysis processes. Understanding the reality that the transaction database is more likely to contain sensitive information such as personal sensitive information, a measure to protect the sensitive information before the data is shared among parties is becoming more crucial. The sensitive information can be leaked by an attacker using an inference attack technique. An approach called a data anonymization scheme that alters an original database to preserve sensitive information can be employed to hinder the attack. Several existing approaches have been suggested to prevent the attack by anonymizing databases. However, it excessively modifies the database in protecting the sensitive information. Consequently, the modified database loses its data utility and changes the database properties significantly. In this paper, we propose a data anonymization method for preserving sensitive information from inference attacks while reducing the amount of data utility loss and maintaining database properties. Experiment evidence suggests that the proposed method outperforms an existing approach which based on global suppression technique i.e., direct algorithm.