Publications

Detailed Information

Multivariate Homomorphic Encryption for Approximate Matrix Arithmetics : 근사 행렬연산을 위한 다변수 동형암호

DC Field Value Language
dc.contributor.advisor천정희-
dc.contributor.author김안드레이-
dc.date.accessioned2019-10-21T03:37:59Z-
dc.date.available2019-10-21T03:37:59Z-
dc.date.issued2019-08-
dc.identifier.other000000156778-
dc.identifier.urihttps://hdl.handle.net/10371/162414-
dc.identifier.urihttp://dcollection.snu.ac.kr/common/orgView/000000156778ko_KR
dc.description학위논문(박사)--서울대학교 대학원 :자연과학대학 수리과학부,2019. 8. 천정희.-
dc.description.abstract혜안(Homomrphic Encryption for Arithmetics of Approximate Numbers, HEAAN) 은 근사 계산을 지원하는 동형 암호 스킴이다. 혜안의 벡터 패킹 기술은 데이터 분석 및 기계 학습 분야 등 근사적인 계산이 필요한 암호화 응용 프로그램에서 효율성을 입증하였다.
다변수 혜안(Multivariate HEAAN, MHEAAN)은 평문의 텐서 구조에 대한 HEAAN의 일반화이다. 본 설계는 연산 과정에서 줄어드는 유효 숫자의 길이가 연 산 서킷의 두께로 제한된다는 HEAAN의 장점을 그대로 가지고, 평문 상태에서의 근사 연산과 비교하였을 때에도 유효 숫자 낭비가 1비트를 넘지 않는다. 평문 벡터 의 회전 등 고차원 벡터의 다양한 구조들이 응용에 많이 쓰임에 따라, MHEAAN은 행렬 및 텐서와 관련된 응용 프로그램에서 기존 HEAAN에 비하여 보다 효율적인 결과를 낳는다.
MHEAAN의 구체적인 2 차원 구조는 행렬 연산에 대한 MHEAAN 기법의 효 율성을 보여 주며, 로지스틱 회귀분석, 심 신경망 구조 및 회귀 신경망 구조와 같은 암호화 된 데이터 및 암호화 된 모델에 대한 여러 기계 학습 알고리즘에 적용될 수 있다. 또한 효율적인 재부팅 구현을 통하여, 이는 임의의 로지스틱 회귀 분석 등의 다양한 응용 분야에 쉽게 활용될 수 있다.
-
dc.description.abstractHomomorphic Encryption for Arithmetics of Approximate Numbers (HEAAN) is a homomorphic encryption (HE) scheme for approximate arithmetics introduced by Cheon et.al. [CKKS17]. Its vector packing technique proved its potential in cryptographic applications requiring approximate computations, including data analysis and machine learning.
Multivariate Homomorphic Encryption for Approximate Matrix Arithmetics (MHEAAN) is a generalization of HEAAN to the case of a tensor structure of plaintext slots. Our design takes advantage of the HEAAN scheme, that the precision losses during the evaluation are limited by the depth of the circuit, and it exceeds no more than one bit compared to unencrypted approximate arithmetics, such as floating-point operations. Due to the multi-dimensional structure of plaintext slots along with rotations in various dimensions, MHEAAN is a more natural choice for applications involving matrices and tensors.
i
The concrete two-dimensional construction shows the efficiency of the MHEAAN scheme on matrix operations and was applied to several Machine Learning algorithms on encrypted data and encrypted model such as Logistic Regression (LR) training algorithm, Deep Neural Network (DNN) and Recurrent Neural Network (RNN) classification algorithms. With the efficient bootstrapping, the implementation can be easily be scaled to the case of arbitrary LR, DNN or RNN structures.
-
dc.description.tableofcontentsAbstract
1 Introduction
1.1 MultidimensionalVariantofHEAAN
1.2 ApplicationsToMachineLearning
1.3 ListOfPapers
2 Background Theory
2.1 BasicNotations
2.2 MachineLearningAlgorithms
LogisticRegression
2.2.2 DeepLearning
2.3 The Cyclotomic Ring and Canonical Embedding
2.4 m-RLWEProblem
2.5 HEAANScheme
2.5.1 BootstrappingforHEAAN
3 MHEAAN Scheme
3.1 MHEAANScheme
3.1.1 StructureofMHEAAN
3.1.2 ConcreteConstruction
3.2 BootstrappingforMHEAAN
3.3 Homomorphic Evaluations of Matrix Operations
3.3.1 MatrixbyVectorMultiplication
3.3.2 MatrixMultiplication
3.3.3 MatrixTransposition
3.3.4 MatrixInverse
4 Applications
4.1 Sigmoid&TanhApproximations
4.2 HomomorphicLRTrainingPhase
4.2.1 DatabaseEncoding
4.2.2 Homomorphic Evaluation of the GD
4.2.3 HomomorphicEvaluationofNLGD
4.3 HomomorphicDNNClassification
4.4 HomomorphicRNNClassification
5 Implementation Results
5.1 EvaluationofNLGDTraining
5.2 EvaluationofDNNClassification
5.3 EvaluationofRNNClassification
6 Conclusions
A Proofs
Abstract (in Korean)
Acknowledgement (in Korean)
-
dc.language.isoeng-
dc.publisher서울대학교 대학원-
dc.subjectCryptography-
dc.subjectHomomorphic Encryption-
dc.subject.ddc510-
dc.titleMultivariate Homomorphic Encryption for Approximate Matrix Arithmetics-
dc.title.alternative근사 행렬연산을 위한 다변수 동형암호-
dc.typeThesis-
dc.typeDissertation-
dc.contributor.department자연과학대학 수리과학부-
dc.description.degreeDoctor-
dc.date.awarded2019-08-
dc.contributor.majorCryptography-
dc.identifier.uciI804:11032-000000156778-
dc.identifier.holdings000000000040▲000000000041▲000000156778▲-
Appears in Collections:
Files in This Item:

Altmetrics

Item View & Download Count

  • mendeley

Items in S-Space are protected by copyright, with all rights reserved, unless otherwise indicated.

Share