• Journal of Information Processing Systems
• /
• v.13 no.5
• /
• pp.1345-1357
• /
• 2017

A prototype selection method chooses a small set of training points from a whole set of class data. As the data size increases, the selected prototypes play a significant role in covering class regions and learning a discriminate rule. This paper discusses the methods for selecting prototypes in a classification framework. We formulate a prototype selection problem into a set covering optimization problem in which the sets are composed with distance metric and predefined classes. The formulation of our problem makes us draw attention only to prototypes per class, not considering the other class points. A training point becomes a prototype by checking the number of neighbors and whether it is preselected. In this setting, we propose a greedy algorithm which chooses the most relevant points for preserving the class dominant regions. The proposed method is simple to implement, does not have parameters to adapt, and achieves better or comparable results on both artificial and real-world problems.

• Journal of the Korea Society of Computer and Information
• /
• v.24 no.4
• /
• pp.45-50
• /
• 2019

Recently, character motion have been used extensively in the entertainment business, and researchers have investigated algorithms of reproducing, editing, and simulating mimic human movements. Also, many recent researches have suggested how a character interacts with its surrounding environment in terms of motion. Specially, spatial relationships of the environment have been introduced for adapting and preserving character motion. In this paper, we propose a motion adaptation technique preserving a spatial property between a virtual character and the configuration of its surrounding space. Additionally, we report on experimental results of smoothly adapted motions in various environmental structures with original motions such as walk, jump, and tumbling.

• Annual Conference of KIPS
• /
• 2024.05a
• /
• pp.229-230
• /
• 2024

최근 드론으로 극한 환경에서 범죄 수배자 및 실종자를 탐색하는 시도가 활발하다. 이때 생체 인증 기술인 얼굴 인증 기술을 사용하면 탐색 효율이 높아지지만, 암호화되지 않은 인증 프로토콜 적용 시 생체 정보 유출의 위험이 있다. 본 논문에서는 드론이 수집한 얼굴 이미지 템플릿을 암호화하여 안전하게 인증할 수 있는 효율적인 생체 인증 프레임워크인 DF-PPHDM(Privacy-Preserving Hamming Distance biometric Matching for Drone-collected Facial images)을 제안한다. 수집된 얼굴 이미지는 암호문 형태로 서버에 전달되며 서버는 기존 등록된 암호화된 템플릿과의 Hamming distance 분석을 통해 검증한다. 제안한 DF-PPHDM을 RaspberryPI 4B 환경에서 직접 실험하여 분석한 결과, 한정된 리소스를 소유한 드론에서 효율적인 구현이 가능하며, 인증 단계에서 7.83~155.03 ㎲ (microseconds)가 소요된다는 것을 입증하였다. 더불어 서버는 드론이 전송한 암호문으로부터 생체 정보를 복구할 수 없으므로 프라이버시 침해 문제를 예방할 수 있다. 향후 DF-PPHDM에 AI(Artificial Intelligence)를 결합하여 자동화 기능을 추가하고 코드 최적화를 통해 성능을 향상시킬 예정이다.

• Proceedings of the IEEK Conference
• /
• 2006.06a
• /
• pp.461-462
• /
• 2006

• International journal of advanced smart convergence
• /
• v.12 no.2
• /
• pp.56-66
• /
• 2023

The goal of deep learning is to extract complex features from multidimensional data use the features to create models that connect input and output. Deep learning is a process of learning nonlinear features and functions from complex data, and the user data that is employed to train deep learning models has become the focus of privacy concerns. Companies that collect user's sensitive personal information, such as users' images and voices, own this data for indefinite period of times. Users cannot delete their personal information, and they cannot limit the purposes for which the data is used. The study has designed a deep learning method that employs privacy protection technology that uses distributed collaborative learning so that multiple participants can use neural network models collaboratively without sharing the input datasets. To prevent direct leaks of personal information, participants are not shown the training datasets during the model training process, unlike traditional deep learning so that the personal information in the data can be protected. The study used a method that can selectively share subsets via an optimization algorithm that is based on modified distributed stochastic gradient descent, and the result showed that it was possible to learn with improved learning accuracy while protecting personal information.

• Journal of KIISE
• /
• v.44 no.8
• /
• pp.860-869
• /
• 2017

Recently, due to the personal information security act, the encryption of personal information has attracted attention. However, if the conventional encryption scheme is used directly, the database schema must be changed because the conventional encryption scheme does not preserve the format of the data, which can yield a large cost. Therefore, the Format-Preserving Encryption(FPE) has emerged as an important technique that ensures the confidentiality of the data and maintains the database schema naturally. Accordingly, National Institute of Standards and Technology(NIST) recently published the FF1 and FF3 as standards for FPE, although problems have been found in the security of FF1 and FF3 against message recovery attacks. In this paper, we study and analyze FF1 and FF3 as the standards of FPE, as well as the message recovery attack on these schemes. We also study a secure FPE against message recovery attack and verify the efficiency by implementing standardized FF1 and FF3.

• Journal of the Korea Society of Computer and Information
• /
• v.26 no.12
• /
• pp.19-27
• /
• 2021

In this paper, we present a method for effectively predicting traffic volume based on vehicle location data that are collected by using LDP (Local Differential Privacy). The proposed solution in this paper consists of two phases: the process of collecting vehicle location data in a privacy-presering manner and the process of predicting traffic volume using the collected location data. In the first phase, the vehicle's location data is collected by using LDP to prevent privacy issues that may arise during the data collection process. LDP adds random noise to the original data when collecting data to prevent the data owner's sensitive information from being exposed to the outside. This allows the collection of vehicle location data, while preserving the driver's privacy. In the second phase, the traffic volume is predicted by applying deep learning techniques to the data collected in the first stage. Experimental results with real data sets demonstrate that the method proposed in this paper can effectively predict the traffic volume using the location data that are collected in a privacy-preserving manner.

• Journal of the Institute of Electronics Engineers of Korea SC
• /
• v.39 no.1
• /
• pp.33-41
• /
• 2002

In this paper, we present a frequency weighted model reduction using LMIs for minimizing the H$\infty$ weighted model error compared with the methods of frequency weighted balanced truncation and frequency weighted Hankel norm approximation. The proposed algorithm, its form is equal to the sufficient condition of performance preserving controller approximation, is based on an iterative two-step LMI scheme induced from bound real lemma. So, it can be applied to the problem of the performance preserving controller approximation. The controller reduction is useful in a practical control design and ensures its easy implementation and high reliability The validity of the proposed algorithm is shown through numerical examples. Additionaly, we extend the proposed algorithm to performance preserving controller approximation by applying to the HIMAT(highly maneuverable aircraft technology) system.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.41 no.6
• /
• pp.25-32
• /
• 2004

We propose a hierarchical disparity estimation algorithm with edge-preserving energy-based regularization. Initial disparity vectors are obtained from downsampled stereo images using a feature-based region-dividing disparity estimation technique. Dense disparities are estimated from these initial vectors with shape-adaptive windows in full resolution images. Finally, the vector fields are regularized with the minimization of the energy functional which considers both fidelity and smoothness of the fields. The first two steps provide highly reliable disparity vectors, so that local minimum problem can be avoided in regularization step. The proposed algorithm generates accurate disparity map which is smooth inside objects while preserving its discontinuities in boundaries. Experimental results are presented to illustrate the capabilities of the proposed disparity estimation technique.

• Journal of Korea Multimedia Society
• /
• v.19 no.7
• /
• pp.1159-1165
• /
• 2016

Traffic violations such as moving while the traffic lights are red have come from a simple omission to a premeditated act. The traffic control center cannot timely monitor all the cameras installed on the roads to trace and pursue those traffic violators. Modern vehicles are equipped and controlled by several sensors in order to support monitoring and reporting those kind of behaviors which some time end up in severe causalities. However, such applications within the vehicle environment need to provide security guaranties. In this paper, we address the limitation of previous work and present a secure and privacy preserving protocol for traffic violation reporting system in vehicular cloud environment which enables the vehicles to report the traffic violators, thus the roadside clouds collect those information which can be used as evidence to pursue the traffic violators. Particularly, we provide the unlinkability security property within the proposed protocol which also offers lightweight computational overhead compared to previous protocol. We consider the concept of conditional privacy preserving authentication without pairing operations to provide security and privacy for the reporting vehicles.