• Journal of the Korean institute of surface engineering
• /
• v.43 no.4
• /
• pp.187-193
• /
• 2010

The colored films formed on 304 stainless steel plates by immersion treatment and electrochemical one in a solution containing sulphuric acids and chromic acids were studied by SEM, AES, and spectrophotometer. The thicknesses of the films by 20 minutes treatment were about 200 nm and it became thinner as the treatment times were increased. The surface texture showed a tortuous network of interlinking pathways. The color of the 304 steel surface changed from metallic white to gray, black, red, yellow-green, and green-blue, gradually, by the treatment time was increased. The reflectivity measured by UV-VIS-NIR spectrophotometer was reduced from max. 38% of basis metal to min. 3.5% of colored surface.

• Proceedings of the KIEE Conference
• /
• 1995.07b
• /
• pp.974-976
• /
• 1995

Segmentation of image data is an important problem in computer vision, remote sensing, and image analysis. Most objects in the real world have textured surfaces. Segmentation based on texture information is possible even if there are no apparent intensity edges between the different regions. There are many existing methods for texture segmentation and classification, based on different types of statistics that can be obtained from the gray-level images. In this paper, we use a neural network model --- ART-2 (Adaptive Resonance Theory) for textures in an image, proposed by Carpenter and Grossberg. In our experiments, we use Walsh matrix as feature value for textured image.

• Current Optics and Photonics
• /
• v.5 no.2
• /
• pp.155-163
• /
• 2021

We propose a novel optical encryption scheme for cipher-feedback-block (CFB) mode, capable of encrypting two-dimensional (2D) page data with the use of two-step phase-shifting digital interferometry utilizing orthogonal polarization, in which the CFB algorithm is modified into an optical method to enhance security. The encryption is performed in the Fourier domain to record interferograms on charge-coupled devices (CCD)s with 256 quantized gray levels. A page of plaintext is encrypted into digital interferograms of ciphertexts, which are transmitted over a digital information network and then can be decrypted by digital computation according to the given CFB algorithm. The encryption key used in the decryption procedure and the plaintext are reconstructed by dual phase-shifting interferometry, providing high security in the cryptosystem. Also, each plaintext is sequentially encrypted using different encryption keys. The random-phase mask attached to the plaintext provides resistance against possible attacks. The feasibility and reliability of the proposed CFB method are verified and analyzed with numerical simulations.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2021.05a
• /
• pp.212-215
• /
• 2021

본 논문에서는 딥러닝의 CNN(Convolution Neural Network) 학습을 통하여 악성코드를 실행시키지 않고서 악성코드 변종을 패밀리 그룹으로 분류하는 방법을 연구한다. 먼저 데이터 전처리를 통해 3가지의 서로 다른 방법으로 악성코드 이미지와 메타데이터를 생성하고 이를 CNN으로 학습시킨다. 첫째, 악성코드의 byte 파일을 8비트 gray-scale 이미지로 시각화하는 방법이다. 둘째, 악성코드 asm 파일의 opcode sequence 정보를 추출하고 이를 이미지로 변환하는 방법이다. 셋째, 악성코드 이미지와 메타데이터를 결합하여 분류에 적용하는 방법이다. 이미지 특징 추출을 위해서는 본고에서 제안한 CNN을 통한 학습 방식과 더불어 3개의 Pre-trained된 CNN 모델을 (InceptionV3, Densnet, Resnet-50) 사용하여 전이학습을 진행한다. 전이학습 시에는 마지막 분류 레이어층에서 본 논문에서 선택한 데이터셋에 대해서만 학습하도록 파인튜닝하였다. 결과적으로 가공된 악성코드 데이터를 적용하여 9개의 악성코드 패밀리로 분류하고 예측 정확도를 측정해 비교 분석한다.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2020.11a
• /
• pp.347-350
• /
• 2020

본 논문에서는 영상 변환 기술인 이산웨이블릿변환(Discrete Wavelet Transform, DWT)를 딥러닝 기반의 네트워크로 구현한다. 딥러닝 기술 중에도 CNN 기반으로 네트워크를 설계하였으며, 본 DWT 네트워크는 해상도에 의존적이지 않은 계층들로만 구성된다. 데이터세트를 구성할 때 파이썬의 라이브러리를 사용하여 레이블 데이터세트를 구성한다. 128×128크기의 gray-scale 영상을 입력으로 사용하고 이에 대응하는 레이블 데이터세트를 구성하여 1-level DWT를 수행하는 네트워크의 학습을 진행한다. 역방향 변환도 네트워크 설계 후 데이터세트를 구성하여 학습을 진행한다. 학습이 완료된 1-level DWT 네트워크를 반복적으로 사용하여 Multi-level DWT 네트워크를 구성한다. 또한 양자화에 의한 간단한 영상압축 실험을 진행하여 DWT 네트워크의 성능과 압축 등의 응용분야에 활용할 수 있음을 보인다. 설계한 DWT 네트워크의 1-level 순방향 변환 성능은 42.18dB의 PSNR을 보였고, 1-level 역방향 변환 성능은 50.13dB의 PSNR을 보였다.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.15 no.3
• /
• pp.1100-1118
• /
• 2021

In a wide range of ML applications, the training data contains privacy-sensitive information that should be kept secure. Training the ML systems by privacy-sensitive data makes the ML model inherent to the data. As the structure of the model has been fine-tuned by training data, the model can be abused for accessing the data by the estimation in a reverse process called model inversion attack (MIA). Although, MIA has been applied to shallow neural network models of recognizers in literature and its threat in privacy violation has been approved, in the case of a deep learning (DL) model, its efficiency was under question. It was due to the complexity of a DL model structure, big number of DL model parameters, the huge size of training data, big number of registered users to a DL model and thereof big number of class labels. This research work first analyses the possibility of MIA on a deep learning model of a recognition system, namely a face recognizer. Second, despite the conventional MIA under the white box scenario of having partial access to the users' non-sensitive information in addition to the model structure, the MIA is implemented on a deep face recognition system by just having the model structure and parameters but not any user information. In this aspect, it is under a semi-white box scenario or in other words a gray-box scenario. The experimental results in targeting five registered users of a CNN-based face recognition system approve the possibility of regeneration of users' face images even for a deep model by MIA under a gray box scenario. Although, for some images the evaluation recognition score is low and the generated images are not easily recognizable, but for some other images the score is high and facial features of the targeted identities are observable. The objective and subjective evaluations demonstrate that privacy cyber-attack by MIA on a deep recognition system not only is feasible but also is a serious threat with increasing alert state in the future as there is considerable potential for integration more advanced ML techniques to MIA.

• Asian Pacific Journal of Cancer Prevention
• /
• v.15 no.12
• /
• pp.4795-4800
• /
• 2014

Background: Few studies have directly compared clinical efficacy and safety among Chinese herb injections (CHIs) for gastric cancer (GC). The present study aimed to compare CHIs combined with FOLFOX regimens for GC to show which provides the best CHIs results. Materials and Methods: 9 electronic databases and 6 gray literature databases were comprehensive searched in April 20, 2013. According to inclusion and exclusion criteria, two reviewers independently selected and assessed the included trials. The risk of bias tool described in the Cochrane Handbook version 5.1.0 and CONSORT statement were used to assess the quality of the trials. All calculations and graphs were performed and produced using ADDIS 1.16.5 software. Results: A total of 541 records were searched and 38 RCTs met the inclusion criteria (2,761 participants), involving 10 CHIs. The results of network meta-analysis showed that compared with FOLFOX alone, combinations with Kanglaite, Astragalus polysaccharides, Cinobufacini, or Yadanziyouru injections could furthest strengthen ORR, improve the quality of life, reduce nausea and vomiting, and reduce the incidence of leukopenia (III-IV). Conclusions: Kanglaite injection, Astragalus polysaccharides injection, Yadanziyouru injection were superior to other CHIs in clinical efficacy and safety for GC. The conclusions now need to be confirmed by large sample size direct head-to-head studies.

• Korean Journal of Biological Psychiatry
• /
• v.24 no.2
• /
• pp.68-74
• /
• 2017

Objectives It is well known that problem drinking is associated with alterations of brain structures and functions. Brain functions related to alcohol consumption can be determined by the resting state functional connectivity in various resting state networks (RSNs). This study aims to ascertain the alcohol effect on the structures forming predetermined RSNs by assessing their cortical thickness. Methods Twenty-six abstinent male patients with alcohol dependence and the same number of age-matched healthy control were recruited from an inpatient mental hospital and community. All participants underwent a 3T MRI scan. Averaged cortical thickness of areas constituting 7 RSNs were determined by using FreeSurfer with Yeo atlas derived from cortical parcellation estimated by intrinsic functional connectivity. Results There were significant group differences of mean cortical thicknesses (Cohen's d, corrected p) in ventral attention (1.01, < 0.01), dorsal attention (0.93, 0.01), somatomotor (0.90, 0.01), and visual (0.88, 0.02) networks. We could not find significant group differences in the default mode network. There were also significant group differences of gray matter volumes corrected by head size across the all networks. However, there were no group differences of surface area in each network. Conclusions There are differences in degree and pattern of structural recovery after abstinence across areas forming RSNs. Considering the previous observation that group differences of functional connectivity were significant only in networks related to task-positive networks such as dorsal attention and cognitive control networks, we can explain recovery pattern of cognition and emotion related to the default mode network and the mechanisms for craving and relapse associated with task-positive networks.

• Fashion & Textile Research Journal
• /
• v.17 no.3
• /
• pp.351-363
• /
• 2015

This research explores diachronic fashion design trends through a structure analysis of a connection network based on fashion show concepts implemented by modern fashion designers from S/S 2009 to F/W 2013. The findings from this research are as follows. First, four categories affect the inspiration and thinking of design: the experience of the designer, social atmosphere and phenomena, natural sensitivity (or formative characteristics of natural objects), and the influence and quality of other fields. Second, in cases where the designers' experiences influenced design inspiration and thinking, designers express personal memories with keywords like high school, grandmother's closet, prom, beauty and the beast, heritage, past, and reminiscence through design elements such as lines, silhouettes, materials, and colors. Third, the representative example of the social atmosphere and phenomena that influenced design inspiration and thinking was the 2008 Global Financial Crisis that reflected the social climate through design concepts of keywords such as Recession, Black, Economy, US, Depression, Gray, Dark Age, White and New York. Fourth, inspired by nature and the formative characteristics in design, the designers employed ornamental elements to collections and design concepts that focused on nature words connected to light, sun, wild, dirt, rock, moss, and trees. Fifth, the designers took their ideas from different fields of personal interest in the arts, science and humanities (sports, architecture, sculpture, painting, and literature) that were decisive in determining materials, design colors and silhouettes. The theme of architecture was analyzed as a central element that had an ongoing impact on the concepts of designers.

• Journal of the Institute of Electronics Engineers of Korea TE
• /
• v.37 no.3
• /
• pp.104-113
• /
• 2000

In this study, we propose facial image recognition based on wavelet transform and neural network. This algorithm is proposed by following processes. First, two gray level images is captured in constant illumination and, after removing input image noise using a gaussian filter, differential image is obtained between background and face input image, and this image has a process of erosion and dilation. Second, a mask is made from dilation image and background and facial image is divided by projecting the mask into face input image Then, characteristic area of square shape that consists of eyes, a nose, a mouth, eyebrows and cheeks is detected by searching the edge of divided face image. Finally, after characteristic vectors are extracted from performing discrete wavelet transform(DWT) of this characteristic area and is normalized, normalized vectors become neural network input vectors. And recognition processing is performed based on neural network learning. Simulation results show recognition rate of 100 % about learned image and 92% about unlearned image.