Privacy-Preserving Naive Bayes Classification Using Fully Homomorphic Encryption

• Published in: Neural Information Processing Vol. 11304; pp. 349 - 358
• Main Authors: Kim, Sangwook, Omori, Masahiro, Hayashi, Takuya, Omori, Toshiaki, Wang, Lihua, Ozawa, Seiichi
• Format: Book Chapter
• Language: English
• Published: Switzerland Springer International Publishing AG 2018; Springer International Publishing
• Series: Lecture Notes in Computer Science
• Subjects: Classification; Fully homomorphic encryption; Machine learning; Naive Bayes; Privacy-preserving data mining
• ISBN: 9783030042110; 3030042111
• ISSN: 0302-9743; 1611-3349
• DOI: 10.1007/978-3-030-04212-7_30

Many services for data analysis require customer’s data to be exposed and privacy issues are critical in related fields. To address this problem, we propose a Privacy-Preserving Naive Bayes classifier (PP-NBC) model which provides classification results without leaking privacy information in data sources. Through classification process in PP-NBC, the operations are evaluated using encrypted data by applying fully homomorphic encryption scheme so that service providers are able to handle customer’s data without knowing their actual values. The proposed method is implemented with a homomorphic encryption library called HElib and we carry out a primitive performance evaluation for the proposed PP-NBC.