公开(公告)号：US10303893B2

公开(公告)日：2019-05-28

申请号：US15366129

申请日：2016-12-01

Applicant: Panasonic Intellectual Property Corporation of America

Inventor： Yuji Unagami ,  Natsume Matsuzaki ,  Shota Yamada ,  Nuttapong Attrapadung ,  Takahiro Matsuda ,  Goichiro Hanaoka

IPC: G06F21/00 ,  G06F21/62 ,  G06F12/14 ,  H04L29/06 ,  G09C1/00 ,  H04L9/00 ,  H04L9/30

Abstract: A data search method of a first device storing multiple sets of privacy data acquired from multiple persons and multiple reference features corresponding to the multiple sets of privacy data, where the multiple reference features each are expressed by an n-dimensional vector, includes receiving first encrypted features from a second device connected to the first device, generating multiple second converted features by a second conversion of the multiple reference features, generating of multiple second encrypted features by encrypting the multiple second converted features using inner product encryption, acquiring multiple inner product values by performing inner product computation of each of the first encrypted features and the multiple second encrypted features, determining whether or not the first features and the first reference features are similar, and transmitting of first privacy data corresponding to the first reference features out of the multiple sets of privacy data to the second device.

公开(公告)号：US11436471B2

公开(公告)日：2022-09-06

申请号：US16149254

申请日：2018-10-02

Applicant: Panasonic Intellectual Property Corporation of America

Inventor： Naohisa Nishida ,  Yuji Unagami ,  Tatsumi Oba ,  Ryo Kato ,  Shota Yamada ,  Nuttapong Attrapadung ,  Tadanori Teruya ,  Takahiro Matsuda ,  Goichiro Hanaoka

IPC: G06N3/04 ,  H04L9/08 ,  G06N3/08

Abstract: A method of obtaining a shared prediction model is provided. The method includes: obtaining a prediction model as a neural network; converting each negative numerical value in a plurality of parameters included in the prediction model to a positive numerical value to obtain a converted prediction model; and sharing the converted prediction model by a secret sharing method to obtain shared prediction models while concealing an input data.

公开(公告)号：US10649919B2

公开(公告)日：2020-05-12

申请号：US15848004

申请日：2017-12-20

Applicant: Panasonic Intellectual Property Corporation of America

Inventor： Yuji Unagami ,  Naohisa Nishida ,  Shota Yamada ,  Nuttapong Attrapadung ,  Takahiro Matsuda ,  Goichiro Hanaoka

IPC: H04L29/06 ,  G06F12/14 ,  H04L9/32 ,  G06F21/40 ,  H04L9/00

Abstract: In an information processing method, a query including a first encrypted feature value provided with confidential information unique to a user is received. The first encrypted feature value is generated by encrypting a first feature value calculated from privacy data of the user by using inner product encryption. A plurality of inner product values are acquired by computing an inner product of the first encrypted feature value and each of a plurality of second encrypted feature values. Privacy data of a plurality of pieces of privacy data having an inner product value of the first encrypted feature value and a second encrypted feature value with an encrypted reference feature value calculated from the privacy data being equal to or smaller than a predetermined threshold is transmitted. A secret key of the user is identified by using the confidential information when an unauthorized access is detected, and identification information is outputted.