分野別セミナー

Privacy preserving data mining (PPDM) is crucially important technology for balancing the data privacy and
the data utility. One strategy for realizing the PPDM is utilizing homomorphic encryption (HE), which enables
us to compute some operations on encrypted data. However, current HEs have several issues, e.g.,
large computational resource requirement and limitation of computable operations. In this talk, firstly we will
introduce our results of HE: security updatable homomorphic encryption and efficient secure inner product
computation, then introduce how to utilize HE to compute logistic regression efficiently and securely.
This is a joint work with my colleagues at NICT: Yoshinori Aono, Le Trieu Phong, and Lihua Wang.