Application of Data Anonymization to Smartphone Apps Usage Logs

In recent years, the importance of data science has rapidly increased, along with opportunities for gaining new knowledge, such as through competitions. This research examines a method for anonymizing log data, using smartphone apps usage logs as an example, that maintains the value of the data while ensuring both anonymity and analyzability. We described a method for generating anonymized data by solving an assignment problem that that shuffles the us-age apps, incorporating time series information, based on Pk-anonymization focused on suites of apps. And we explored three anonymizing methods. Our verification using smartphone apps usage logs data for cluster analysis shows that all methods effectively replicate trend analysis. The method generating data to shuffle fully was found to be overprocessed, and considering time series information proved effective. Comparing two methods with and without Pk-annymization, it offered more robustness but slightly affect the accuracy of detailed values in cases of fewer logs. However without affecting the accuracy of cluster classification. Pk-anonymization was confirmed useful for maintaining confidential-ity in third-party data provision or avoiding identification from niche app usage. Therefore, the method with Pk-anonymization is recommended for situations requiring anonymity, and one without Pk-anonymization for scenarios prioritizing analysis potential, including time series log analysis. Future work are improving methods to better carryover time series information and to more accurately address cases with fewer logs, bringing detailed values closer to the original.