Abstract

According to the recent amendment of data-related bills, active efforts are being made to conduct big data analysis for hyper-personalization services. Hyper-personalization services involve several data integration, including external data. However, it also inherently possesses limitations in usage due to concerns about private information leaks, which can lead to damages to individuals and businesses and unnecessary societal losses. One approach to mitigate such security risks is to use homomorphic encryption, which protects private information. In particular, homomorphic encryption fundamentally prevents private information leakage during data integration and analysis processes and retains original data. In this study, we develop an algorithm for performing inference using an ensemble learning method with homomorphic encryption and apply it to real financial complaint data. We demonstrate the equivalence between scores obtained from random forest and XGBoost in plaintext and those derived from the homomorphic ciphertext. Empirical results show that the algorithm achieves inference computation times in approximately 0.4 seconds, confirming the potential utility of the tree-based ensemble learning method.

Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call