PPS-FPCM: PRIVACY-PRESERVING SEMI-FUZZY POSSIBILISTIC C-MEANS

Applying traditional clustering techniques to big data on the cloud while preserving privacy of is a challenge due required division and exponential operations in each iteration, which complicate its implementation encrypted data. Several existing approaches are based approximating formulas centers, weights, memberships as three polynomial functions according multivariate Taylor formula. However, they usually suffer an increase complexity slight drop accuracy. In this paper, novel Privacy-Preserving semi-fuzzy algorithm possibilistic paradigm, termed PPS-FPCM, presented. Its main feature that it avoids exponentiation operations, at without losing By restricting typicality ordered set discrete values between zero one decided by owner (DO), computation simplified. The second key idea use soft detect outliers compute corresponding memberships, used in-between cluster distance. initial requires magnitude relation comparison, still difficult do over research study, we show how incomplete re-encryption method can be tackle problem. centers distances new computed calculator server (CaCS) responsible for storing cipher texts (DO)’s processing them. Then, CaCS sends incompletely re-encrypted difference these iteratively updated bin correspond initially (DO) comparator (CoCS). CoCS decrypts returns results comparisons. When receives comparison from CoCS, decides appropriate or resends same distance another value. number comparisons O(log bins). computes refined updates centers. end, final (DO). proposed applicable normal homomorphically data, more effective than several related algorithms, produce accurate using large enough (16 more) with high reduction runtime much less complex added communication cost CoCS.