Consideration of Privacy-Preserving Data Mining for Social Networking Service Data with Partially Anonymized

The Chapter One presents the issue of how to consider anonymization methods for datasets that contain publicly available information, such as SNS data and confidential information. Based on the experimental results, it has been confirmed that the nonanonymized attributes include those that render the narrowing down of data easy and those that make it otherwise. This indicates that there are attributes that cannot be anonymized to increase the degree of security. Since attributes that make it easy to narrow down data contain data with few distributions, it is safer to delete such attributes as identifiers when using the anonymization approach. As a conclusion of this study, the anonymization approach can be applied to partially anonymized data, such as SNS data. However, it is necessary to consider the deletion of the attribute as an identifier, depending on the data distribution.