• Journal of the Korea Institute of Information Security & Cryptology
• /
• v.30 no.6
• /
• pp.1291-1300
• /
• 2020

Recently, a deep learning-based non-profiling side-channel analysis was proposed. The deep learning-based non-profiling analysis is a technique that trains a neural network model for all guessed keys and then finds the correct secret key through the difference in the training metrics. As the performance of non-profiling analysis varies greatly depending on the neural network training model design, a correct model design criterion is required. This paper describes the two types of loss functions and eight labeling methods used in the training model design. It predicts the analysis performance of each labeling method in terms of non-profiling analysis and power consumption model. Considering the characteristics of non-profiling analysis and the HW (Hamming Weight) power consumption model is assumed, we predict that the learning model applying the HW label without One-hot encoding and the Correlation Optimization (CO) loss will have the best analysis performance. And we performed actual analysis on three data sets that are Subbytes operation part of AES-128 1 round. We verified our prediction by non-profiling analyzing two data sets with a total 16 of MLP-based model, which we describe.

• The Journal of the Acoustical Society of Korea
• /
• v.40 no.5
• /
• pp.530-536
• /
• 2021

Recently, due to the development of deep learning, end-to-end speech recognition, which directly maps graphemes to speech signals, shows good performance. Especially, among the end-to-end models, conformer shows the best performance. However end-to-end models only focuses on the probability of which grapheme will appear at the time. The decoding process uses a greedy search or beam search. This decoding method is easily affected by the final probability output by the model. In addition, the end-to-end models cannot use external pronunciation and language information due to structual problem. Therefore, in this paper conformer with lexicon transducer is proposed. We compare phoneme-based model with lexicon transducer and grapheme-based model with beam search. Test set is consist of words that do not appear in training data. The grapheme-based conformer with beam search shows 3.8 % of CER. The phoneme-based conformer with lexicon transducer shows 3.4 % of CER.

• Smart Media Journal
• /
• v.12 no.1
• /
• pp.9-16
• /
• 2023

Recently, a convolutional neural network (CNN) based system is being developed to overcome the limitations of human resources in the apple quality classification of farmhouse. However, since convolutional neural networks receive only images of the same size, preprocessing such as sampling may be required, and in the case of oversampling, information loss of the original image such as image quality degradation and blurring occurs. In this paper, in order to minimize the above problem, to generate a image patch based graph of an original image and propose a random walk-based positional encoding method to apply the graph transformer model. The above method continuously learns the position embedding information of patches which don't have a positional information based on the random walk algorithm, and finds the optimal graph structure by aggregating useful node information through the self-attention technique of graph transformer model. Therefore, it is robust and shows good performance even in a new graph structure of random node order and an arbitrary graph structure according to the location of an object in an image. As a result, when experimented with 5 apple quality datasets, the learning accuracy was higher than other GNN models by a minimum of 1.3% to a maximum of 4.7%, and the number of parameters was 3.59M, which was about 15% less than the 23.52M of the ResNet18 model. Therefore, it shows fast reasoning speed according to the reduction of the amount of computation and proves the effect.

• Journal of KIISE
• /
• v.44 no.12
• /
• pp.1290-1295
• /
• 2017

Document classification, a task of classifying the category of each document based on text, is one of the fundamental areas for natural language processing. Document classification may be used in various fields such as topic classification and sentiment classification. Neural network models for document classification can be divided into two categories: word-level models and character-level models that treat words and characters as basic units respectively. In this study, we propose a neural network model that combines character-level and word-level models to improve performance of document classification. The proposed model extracts the feature vector of each word by combining information obtained from a word embedding matrix and information encoded by a character-level neural network. Based on feature vectors of words, the model classifies documents with a hierarchical structure wherein recurrent neural networks with attention mechanisms are used for both the word and the sentence levels. Experiments on real life datasets demonstrate effectiveness of our proposed model.

• Journal of the Korea Convergence Society
• /
• v.8 no.10
• /
• pp.167-172
• /
• 2017

The purpose of this study is to analyze the emotional codes of Sijo, which has been acknowledged to have excellent therapeutic function, to activate the contents of the therapy of humanities. Sijo as a function of healing forms emotional codes of therapy, which is the total of emotions, through the fusion of emotions formed during the process of appreciation of various works. This process enables the literary therapeutic activities to proceed physiologically in the human body. Just as machine learning is self-learning by cognitive functions, the coding process for encoding and re-encoding at all times operates on collections of numerous neurons in the human system. In such a process, it is predicted that amino acids are synthesized in the human body by collective encoding of emotion codes. These amino acids regulate the signaling system of the human body. In the future, if the study on the healing process as such at the contact point of humanities and human physiology proceeds, it is expected that a program of higher quality humanistic therapy will be activated.

• Journal of Digital Convergence
• /
• v.15 no.7
• /
• pp.55-64
• /
• 2017

The purpose of this study is to identify the difficulties of learners who started programming after entering college and to search an effective software education method as university liber arts for non-science major students. In order to do this, we analyzed the difficulties of learners in Python programming classes composed of students from various majors at H University through questioning and taught them using flipped class model with pre-questions. The questions that students submit are collected online before class every time, the data on the degree of the difficulty of feeling and the understanding of feeling were obtained through the questionnaire. As a result, for learners who are new to programming, the learners should allocate the process of making the problem into a logical abstraction at the beginning of the curriculum before learning the basic concept of computer language, each lesson should be practiced through the bottom-up problems enough to provide a logical understanding before actual coding. In addition, detailed curriculum should be developed according to characteristics of learner's major, contents and conducting level.

• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.22 no.5
• /
• pp.75-86
• /
• 2022

This study aims to solve the problem of class imbalance in numerical data by using a deep learning-based Variational AutoEncoder and to improve the performance of the learning model by augmenting the learning data. We propose 'D-VAE' to artificially increase the number of records for a given table data. The main features of the proposed technique go through discretization and feature selection in the preprocessing process to optimize the data. In the discretization process, K-means are applied and grouped, and then converted into one-hot vectors by one-hot encoding technique. Subsequently, for memory efficiency, sample data are generated with Variational AutoEncoder using only features that help predict with RFECV among feature selection techniques. To verify the performance of the proposed model, we demonstrate its validity by conducting experiments by data augmentation ratio.

• Journal of the Korea Society of Computer and Information
• /
• v.27 no.5
• /
• pp.149-156
• /
• 2022

Recently, there have been many reports of document-type malicious code injecting malicious code into Microsoft Office files. Document-type malicious code is often hidden by encoding the malicious code in the document. Therefore, document-type malware can easily bypass anti-virus programs. We found that malicious code was inserted into the Visual Basic for Applications (VBA) macro, a function supported by Microsoft Office. Malicious codes such as shellcodes that run external programs and URL-related codes that download files from external URLs were identified. We selected 354 keywords repeatedly appearing in malicious Microsoft Office files and defined the number of times each keyword appears in the body of the document as a feature. We performed machine learning with SVM, naïve Bayes, logistic regression, and random forest algorithms. As a result, each algorithm showed accuracies of 0.994, 0.659, 0.995, and 0.998, respectively.

• Communications of Mathematical Education
• /
• v.34 no.1
• /
• pp.1-15
• /
• 2020

Today's healthcare, intelligent robots, smart home systems, and car sharing are already innovating with cutting-edge information and communication technologies such as Artificial Intelligence (AI), the Internet of Things, the Internet of Intelligent Things, and Big data. It is deeply affecting our lives. In the factory, robots have been working for humans more than several decades (FA, OA), AI doctors are also working in hospitals (Dr. Watson), AI speakers (Giga Genie) and AI assistants (Siri, Bixby, Google Assistant) are working to improve Natural Language Process. Now, in order to understand AI, knowledge of mathematics becomes essential, not a choice. Thus, mathematicians have been given a role in explaining such mathematics that make these things possible behind AI. Therefore, the authors wrote a textbook 'Basic Mathematics for Artificial Intelligence' by arranging the mathematics concepts and tools needed to understand AI and machine learning in one or two semesters, and organized lectures for undergraduate and graduate students of various majors to explore careers in artificial intelligence. In this paper, we share our experience of conducting this class with the full contents in http://matrix.skku.ac.kr/math4ai/.

• Informatization Policy
• /
• v.24 no.4
• /
• pp.3-16
• /
• 2017

With the development of social media services in the era of Web 2.0, the public opinion formation site has been partially shifted from the traditional mass media to social media. This phenomenon is continuing to expand, and public opinions on government polices created and shared on social media are attracting more attention. It is particularly important to grasp public opinions in policy formulation because setting up educational policies involves a variety of stakeholders and conflicts. The purpose of this study is to explore public opinions about education-related policies through an empirical analysis of social media documents on education policies using opinion mining techniques. For this purpose, we collected the education policy-related documents by keyword, which were produced by users through the social media service, tokenized and extracted sentimental qualities of the documents, and scored the qualities using sentiment dictionaries to find out public preferences for specific education policies. As a result, a lot of negative public opinions were found regarding the smart education policies that use the keywords of digital textbooks and e-learning; while the software education policies using coding education and computer thinking as the keywords had more positive opinions. In addition, the general policies having the keywords of free school terms and creative personality education showed more negative public opinions. As much as 20% of the documents were unable to extract sentiments from, signifying that there are still a certain share of blog posts or tweets that do not reflect the writers' opinions.