• Proceedings of the Korea Information Processing Society Conference
• /
• 2021.11a
• /
• pp.313-314
• /
• 2021

최근 개인 정보 보호를 위해 주목 받고 있는 동형암호 알고리즘은 암호화된 상태로 덧셈과 곱셈 연산이 가능하여, 연산을 위한 복호화 과정 없이 데이터에 대한 가공이 가능하다. 따라서 이러한 동형암호 알고리즘이 개인 정보 보호를 위한 방법으로 떠오르고 있으며, 특히 완전동형암호 알고리즘의 경우 덧셈과 곱셈 연산을 모두 지원하며, 유효 연산 횟수에도 제한이 없어 응용 분야에서 널리 활용될 것으로 예상된다. 그러나, 완전동형암호 알고리즘의 경우 암호문의 크기가 평문대비 크게 증가하고, 다항식으로 구성된 암호문의 덧셈 및 곱셈 연산도 복잡하여 이에 대한 가속이 필요한 실정이다. 이에 FPGA 기반의 동형암호 가속기 개발이 많이 연구되고 있으며, 이를 통해 동형암호 연산의 특징을 이해하고 가속기 연구 동향을 알아보려 한다.

• Review of KIISC
• /
• v.30 no.5
• /
• pp.111-119
• /
• 2020

동형암호(homomorphic encryption)는 암호화된 데이터 사이에서 임의의 연산을 가능하게 하는 유망한 암호학적 스킴(scheme)이다. 이를 활용하면 암호화된 데이터를 복호화하지 않고, 암호화된 상태에서 임의의 연산을 수행 할 수 있을 뿐만아니라, 격자를 기반(lattice-based)으로 하여 양자 알고리즘에 내성(resistant)이 있어 안전하다. 하지만, 동형암호를 이해하기 위해서는 전문적인 암호 또는 계산적인 이론의 지식과 이해가 필요하다. 따라서 본 논문에서는 완전동형암호(fully homomorphic encryption)의 기저에 있는 LWE(learning with error) 문제에서부터 완전동형암호의 핵심인 NAND 게이트와 부트스트래핑(bootstrapping)까지의 과정을 어렵지 않게 설명하여 초보자들의 이해를 돕고자 한다.

• The Journal of the Korea Contents Association
• /
• v.13 no.8
• /
• pp.36-43
• /
• 2013

Fully homomorphic encryption is a cryptography system in which coded data can be searched and statistically processed without decryption. Fully homomorphic encryption has accelerated searching speed by minimizing time spent on encryption and decryption. In addition, it is also known to prevent leakage of any data decoded for statistical reasons. Also, it is expected to protect personal information stored in the cloud computing environment which is becoming commercialized. Since the 1970s when fully homomorphic encryption was first introduced, it has been researched to develop the algorithm that satisfy effectiveness and functionality. We will take the reader through a journey of these developments and provide a glimpse of the exciting research directions that lie ahead.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.25 no.12
• /
• pp.1904-1913
• /
• 2021

With the proliferation of smart phones and the development of IoT technology, it has become possible to collect personal data for public purposes. However, users are afraid of voluntarily providing their private data due to privacy issues. To remedy this problem, mainly three techniques have been studied: data disturbance, traditional encryption, and homomorphic encryption. In this work, we perform simulations to compare them in terms of accuracy, message length, and computation delay. Experiment results show that the data disturbance method is fast and inaccurate while the traditional encryption method is accurate and slow. Similar to traditional encryption algorithms, the homomorphic encryption algorithm is relatively effective in privacy preserving because it allows computing encrypted data without decryption, but it requires high computation costs as well. However, its main cost, arithmetic operations, can be processed in parallel. Also, data analysis using the homomorphic encryption needs to do decryption only once at any number of data.

• The Journal of the Korea Contents Association
• /
• v.21 no.9
• /
• pp.42-50
• /
• 2021

As the Internet environment develops, data-analysis-based applications have been widely and extensively used in the past decade. However, these applications potentially have a privacy problem in that users' personal information may be leaked to unauthorized parties. To tackle such a problem, researchers have suggested several techniques including data perturbation and cryptography. The homomorphic encryption algorithm is a relatively new cryptography technology that allows arithmetic operations for encrypted values as it is without decryption. Since original values are not required, we believe that this method provides better privacy protection than other existing solutions. In this work, we propose to apply a homomorphic encryption algorithm that protects personal information while enabling data analysis.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2022.05a
• /
• pp.505-508
• /
• 2022

딥러닝 기술이 발전하면서 적용되는 산업 분야가 늘어남에 따라 딥러닝 모델에서 역으로 학습 데이터를 추출하는 등 다양한 딥러닝 모델 공격 이슈가 발생하고 있다. 이러한 위협에 대응하기 위해 딥러닝 학습에 사용되는 데이터의 노출을 방지할 수 있도록 사용자 프라이버시를 보호하는 기법의 중요성이 대두되고 있다. 동형암호는 학습 데이터를 보호할 수 있는 기법 중 하나로, 복호화 과정없이 암호화된 상태로 연산, 탐색, 분석 등을 수행할 수 있는 차세대 암호 알고리즘이다. 본 논문에서는 동형암호 기반의 딥러닝 기법 연구 동향에 대해 알아본다.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2016.05a
• /
• pp.762-763
• /
• 2016

In realizing a homomorphic encryption system, the operations of encrypt, decypt, and recrypt constitute major portions. The most important common operation for each back-bone operations include a polynomial modulo multiplication for over million-bit integers, which can be obtained by performing integer Fourier transform, also known as number theoretic transform. In this paper, we adopt and modify an algorithm for calculating big integer multiplications introduced by Schonhage-Strassen to propose an efficient algorithm which can save memory. The proposed architecture of number theoretic transform has been implemented on an FPGA and evaluated.

• Journal of the Korea Society of Computer and Information
• /
• v.26 no.4
• /
• pp.1-9
• /
• 2021

In this paper, we propose a training algorithm of support vector machine (SVM) with a sensitive variable. Although machine learning models enable automatic decision making in the real world applications, regulations prohibit sensitive information from being used to protect privacy. In particular, the privacy protection of the legally protected attributes such as race, gender, and disability is compulsory. We present an efficient least square SVM (LSSVM) training algorithm using a fully homomorphic encryption (FHE) to protect a partial sensitive attribute. Our framework posits that data owner has both non-sensitive attributes and a sensitive attribute while machine learning service provider (MLSP) can get non-sensitive attributes and an encrypted sensitive attribute. As a result, data owner can obtain the encrypted model parameters without exposing their sensitive information to MLSP. In the inference phase, both non-sensitive attributes and a sensitive attribute are encrypted, and all computations should be conducted on encrypted domain. Through the experiments on real data, we identify that our proposed method enables to implement privacy-preserving sensitive LSSVM with FHE that has comparable performance with the original LSSVM algorithm. In addition, we demonstrate that the efficient sensitive LSSVM with FHE significantly improves the computational cost with a small degradation of performance.

• Journal of Internet of Things and Convergence
• /
• v.8 no.1
• /
• pp.79-85
• /
• 2022

Internet of Things (IoT) connects devices with various platforms, computing power, and functions. Due to the diversity of networks and the ubiquity of IoT devices, demands for security and privacy are increasing. Therefore, cryptographic mechanisms must be strong enough to meet these increased requirements, while at the same time effective enough to be implemented in devices with long-range specifications. In this paper, we present the performance and memory limitations of modern cryptographic primitives and schemes for different types of devices that can be used in IoT. In addition, detailed performance evaluation of the performance of the most commonly used encryption algorithms in low-spec devices frequently used in IoT networks is performed. To provide data protection, the binary ring uses encryption asymmetric fully homomorphic encryption and symmetric encryption AES 128-bit. As a result of the experiment, it can be seen that the IoT device had sufficient performance to implement a symmetric encryption, but the performance deteriorated in the asymmetric encryption implementation.

• Journal of the Korea Institute of Information Security & Cryptology
• /
• v.28 no.6
• /
• pp.1401-1414
• /
• 2018

The k-means clustering algorithm groups input data with the number of groups represented by variable k. In fact, this algorithm is particularly useful in market segmentation and medical research, suggesting its wide applicability. In this paper, we propose a privacy-preserving clustering algorithm that is appropriate for outsourced encrypted data, while exposing no information about the input data itself. Notably, our proposed model facilitates encryption of all data, which is a large advantage over existing privacy-preserving clustering algorithms which rely on multi-party computation over plaintext data stored on several servers. Our approach compares homomorphically encrypted ciphertexts to measure the distance between input data. Finally, we theoretically prove that our scheme guarantees the security of input data during computation, and also evaluate our communication and computation complexity in detail.