Convolutional neural network model verification method and device

Disclosed are a method for verifying a convolutional neural network model and a device thereof, which can verify identity by generating a stipulated and proof value by separating a convolution operation and other operations in a convolutional neural network model. The method comprises: (a) a step of...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: OH HYUN OK, KIM JI HYE, LEE SEUNG HWA, KO HAN KYUNG
Format: Patent
Sprache:eng ; kor
Schlagworte:
Online-Zugang:Volltext bestellen
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:Disclosed are a method for verifying a convolutional neural network model and a device thereof, which can verify identity by generating a stipulated and proof value by separating a convolution operation and other operations in a convolutional neural network model. The method comprises: (a) a step of generating a polynomial circuit formula for a first configuration among a plurality of configurations constituting the convolutional neural network model; (b) a step of generating a first stipulated value and a first proof value by applying a zero-knowledge proof scheme based on the polynomial circuit formula; (c) a step of generating a calculation circuit formula for a second one of the plurality of configurations; (d) a step of generating a second stipulated value and a second proof value by applying a zero-knowledge proof scheme based on the arithmetic circuit formula; and (e) a step of generating a connection proof value linking the first stipulated value and the second stipulated value. 컨볼루션 신경망 모델 검증 방법 및 그 장치가 개시된다. 컨볼루션 신경망 모델 검증 방법은, (a) 상기 컨볼루션 신경망 모델을 구성하는 복수의 구성 중 제1 구성에 대해 다항식 회로식을 생성하는 단계; (b) 상기 다항식 회로식을 기초로 영지식 증명 스킴을 적용하여 제1 약정값 및 제1 증명값을 생성하는 단계; (c) 상기 복수의 구성 중 제2 구성에 대해 산술 회로식을 생성하는 단계; (d) 상기 산술 회로식을 기초로 영지식 증명 스킴을 적용하여 제2 약정값 및 제2 증명값을 생성하는 단계; 및 (e) 상기 제1 약정값과 상기 제2 약정값을 연결하는 연결 증명값을 생성하는 단계를 포함한다.