eISSN : 3022-7011
ISSUER : KIPS
 
After the Korea Information Processing Society (KIPS) Transactions journal was founded in 1994, it was reorganized into the KIPS Transactions: Computer and Communication Systems(2287-5891/2734-049X ) and the KIPS Transactions: Software and Data Engi neering(2287-5905/2734-0503) in 2012. Through the KIPS official meeting on January 8th, 2024, the new KIPS Transaction journal was founded by integrating two KIPS Journals, KIPS Transactions: Computer and Communication Systems and KIPS Transactions: Software and Data Engineering. The new journal aims to realize social value and contribute to the development of South Korea’s science and technology with support from the lottery fund of the Ministry of Strategy and Finance and the science/technology promotion fund of the Ministry of Science and ICT. It is indexed in the Korea Science Academic Database, Korea Citation Index (KCI), and EBSCO.

HighlightsMore

Performance Evaluation and Consideration of Shadow Stack on RISC-V Architecture

Kang Ha Young  Han Go Won  Park Sung Hwan  Kwon Dong Hyun

RISC-V is an open-source instruction set architecture, used in various hardware implementations, and can be flexibly expanded to meet system requirements through the RV64I base instruction set and 16 standard extensions. Currently, the RISC-V archit...

A Study on the Evaluation Methods for Assessing the Understanding of Korean Culture by Generative AI Models

Son Ki Jun  Kim Seung Hyun

Recently, services utilizing large-scale language models (LLMs) such as GPT-4 and LLaMA have been released, garnering significant attention. These models can respond fluently to various user queries, but their insufficient training on Korean data ra...

Protection on WebAssembly JIT-Compiled Code with Randomized Memory Protection Key

Shin Chae Won  Jeong Yun Seo  Bae Myeong Jin  Kwon Dong Hyun

WebAssembly(Wasm) is a powerful platform that enables compiled code in various programming languages to be executed in web browser and in varied runtime environments. Specifically, for performance optimization, Wasm runtime provides Just-in-Time (JI...

A Study on Automatic Metrics for Korean Text Abstractive Summarization

Sehwi Yoon  Youhyun Shin

This study aims to analyze and validate automatic evaluation metrics for Korean abstractive summarization. The unique linguistic characteristics of each language require evaluation metrics designed for them, underscoring the importance of research f...

Latest Publication   (Vol. 14, No. 5, May.  2025)

Label Differential Privacy Study for Privacy Protection in Multimodal Contrastive Learning Model
Youngseo Kim  Minseo Yu  Younghan Lee  Ho Bae
Recent advancements in multimodal deep learning have garnered significant attention from both academia and industry due to their exceptional accuracy and ability to learn rich knowledge representations. In particular, contrastive learning based approaches have played a pivotal role in dramatically enhancing the performance of multimodal deep learning. However, the use of multiple data sources in multimodal deep learning increases the risk of inferring sensitive information through data fusion, posing a higher privacy invasion attack compared to unimodal deep learning. This challenge cannot be fully addressed by privacy preserving techniques traditionally employed in unimodal deep learning, underscoring the growing importance of privacy protection in this domain. To address this issue, previous studies have relied on trusted execution environments or strengthened security by selectively recording data classified as privacy threatening. However, these approaches face limitations such as hardware dependency, performance degradation, and accuracy issues in data classification. These shortcomings hinder scalability and usability while leaving systems vulnerable to emerging threats. In this study, we address the privacy concerns by applying the Double Randomized Response algorithm, which ensures label differential privacy during the data preparation process. As a result, we achieved 80.14% accuracy in image-table matching and classification tasks, demonstrating a balance between privacy protection and performance. This method is the first to incorporate data security considerations into multimodal deep learning models while substantiating its efficacy, marking a significant contribution to the field.
The Transactions of the Korea Information Processing Society, Vol. 14, No. 5, pp. 289-296, May. 2025
https://doi.org/10.3745/TKIPS.2025.14.5.289
Differential privacy multimodal deep learning contrastive learning data privacy
A Machine Learning Model for Rapid Prediction of Smart Media Addiction Tendencies Based on Survey Data
Hong Seung Hyeon  Kwon Gi Hyuk  Lee Seo Jin  Kim Jongwan
This study proposes an artificial intelligence model that utilizes machine learning to rapidly classify adults into three categories— high-risk, potential risk, and normal—based on big data from a survey of smartphone addiction. Smart media continues to evolve, and its reliance has surged significantly since the COVID-19 pandemic; however, current prevention and treatment programs for smartphone addiction are insufficient. Most smart media users are unaware of their addiction, and even when they recognize the potential risks of addiction, the prediction of addictive tendencies primarily occurs by presenting scales to clients or visitors seeking psychological assessments. As a result, the general public's access to assessments for smartphone addiction scales is notably limited. This study aims to enhance public access to smartphone addiction scales by developing an automated artificial intelligence model using machine learning, in order to identify addiction trends on both individual and group levels. The artificial intelligence model proposed in this study enhances accuracy and processing time by substituting the score processing of traditional scales with machine learning techniques. The results of this study can contribute to real-time recognition and prevention of smartphone addiction by reducing manpower and time costs during the classification process, thereby promoting healthier usage habits through the reduction of addiction risks. In counseling clients, an AI-based approach can be utilized for the development of personalized prevention and treatment programs for addiction. Moreover, the collected data will serve as an important foundational resource for understanding trends in smartphone addiction as big data.
The Transactions of the Korea Information Processing Society, Vol. 14, No. 5, pp. 297-304, May. 2025
https://doi.org/10.3745/TKIPS.2025.14.5.297
Smart Media Smart Media Addiction Machine Learning Artificial intelligence Counseling Psychology
Sound Signal-based Attendance System for Preventing Unauthorized Attendance
Yeonwoo Sea  Jihwan Park  Dayeong Kang  Jiyeon Lee
Attendance is an important indicator that can confirm students’ participation in lectures, and the traditional method of attending by name has the disadvantage of taking a lot of time. To solve this problem, many universities have introduced an electronic attendance system to simplify the attendance authentication process, but the current authentication methods have a problem in that attendance authentication outside the classroom is possible due to information leakage or additional equipment needs to be installed to introduce the system. In this paper, we propose a sound wave signal-based attendance authentication system for the convenience of introducing the system in the future to prevent unauthorized attendance. The proposed system was able to authenticate attendance within an average of about 16 seconds for students up to 11 meters away from the speaker with an output of 55 dB in a 100m² classroom, and it was confirmed that signal leakage using sound transmission methods such as recording and call was impossible.
The Transactions of the Korea Information Processing Society, Vol. 14, No. 5, pp. 305-311, May. 2025
https://doi.org/10.3745/TKIPS.2025.14.5.305
Unauthorized Attendance Sound Signal authentication LMS
Techniques for Visualizing and Designing Singularity Cells and Weight-Based Subgrid Ranges in the Wave Function Collapse Algorithm
JooHyun Bae  Mankyu Sung
This paper proposes a modified method of the Wave Function Collapse (WFC) algorithm to automatically generate essential 2D game maps using given tiles. As a Procedural Content Generation (PCG) technique, the proposed method addresses two major limitations of the original WFC algorithm: the lack of designer intent reflection and slow computational speed. The method generates the map in subgrid units and introduces techniques such as handling singularity cells and controlling weight values. These enhancements improve the algorithm's processing speed and enable the generation of diverse map designs that can satisfy player expectations. Experimental results demonstrate a performance improvement ranging from at least 10 times up to 150 times faster than the original approach.
The Transactions of the Korea Information Processing Society, Vol. 14, No. 5, pp. 312-319, May. 2025
https://doi.org/10.3745/TKIPS.2025.14.5.312
PCG WFC Game weight Game-Design
Hybrid Feature Selection Method for Replay Attack Detection on Lightweight Devices
Gyujeong Jin  Seyoung Lee
As wearable devices with voice assistants, like smartwatches, become more common, the threat of replay attacks—where recorded voice commands are replayed to bypass authentication—continues to grow, posing serious risks to user security. While lightweight detection models have been proposed, their real-world applicability is limited due to the computational burden and redundancy of high-dimensional features. This study presents a lightweight detection framework using a reinforcement learning-based feature selection method that automatically selects only the most relevant features. By combining SHAP, Permutation Importance, and reinforcement learning, a reliable Top-K feature set is built to enhance both explainability and generalizability. The framework is broadly applicable across feature sets and adaptable to various voice-based environments. Experiments show that it maintains high detection performance with fewer features while reducing computational cost and improving training efficiency.
The Transactions of the Korea Information Processing Society, Vol. 14, No. 5, pp. 320-331, May. 2025
https://doi.org/10.3745/TKIPS.2025.14.5.320
Feature selection Replay Attack Detection Lightweight Security Framework
A Metric Selection Framework for Quantitative Evaluation of Cyber Resilience
Hye-Jin Kang  Ji-Hyun Sung  Harksu Cho
This study proposes a framework for selecting quantitative metrics to evaluate cyber resilience. Based on an analysis of various service types certified under information security and personal data protection management systems (ISMS-P), a pool of candidate resilience metrics was derived. The validity of the selected metrics was evaluated according to five key criteria defined in this study: objectivity, reproducibility, scalability, practicality, and resilience representation. In addition, the principles of mutual exclusivity (ME) and collective exhaustiveness (CE) were applied as supplementary criteria to eliminate redundancy and ensure the general applicability of the evaluation framework. As a result, 12 quantitative resilience metrics were selected. Among them, a detailed empirical analysis was conducted on the Transactions Per Second (TPS) metric, which measures transaction throughput in systems such as web services, focusing on its variation and interpretation under TCP SYN flooding attack scenarios. The proposed metric selection framework establishes a standardized evaluation framework for objectively measuring cyber resilience, thereby providing a basis for effectively responding to and enhancing resilience against continuously evolving cyber threats.
The Transactions of the Korea Information Processing Society, Vol. 14, No. 5, pp. 332-342, May. 2025
https://doi.org/10.3745/TKIPS.2025.14.5.332
Cyber Resilience Quantitative Evaluation Performance metrics Information security management system
Self-Supervised Waste Image Classification Model Considering Intra-Class Diversity
Hyuksoon Choi  Jinhwan Yang  Nammee Moon  Jinah Kim
In this paper, we propose ReCLNet(Reconstructed patch Contrastive Learning Network), a self-supervised learning model that integrates MAE(Masked AutoEncoder) and CL(Contrastive Learning) to address the challenge of generalizing waste image classification, which is often hindered by high intra-class visual diversity and complex backgrounds. To resolve the issue of representation interference between reconstruction and contrastive learning observed in previous combined models, ReCLNet utilizes reconstructed patches for contrastive learning and adopts a dual-encoder architecture with shared weights, thereby ensuring both representation alignment and training consistency. Experimental results show that ReCLNet achieves lower loss and higher classification accuracy compared to TMAC(Transformer-based MAE using Contrastive learning), MAE, CL, and supervised learning-based models. Furthermore, it demonstrates robust and generalized representation learning even in environments with high intra-class variability. These results suggest that ReCLNet has strong potential not only for real-world automated waste classification systems but also for a wide range of applications that require the understanding of complex visual information.
The Transactions of the Korea Information Processing Society, Vol. 14, No. 5, pp. 343-351, May. 2025
https://doi.org/10.3745/TKIPS.2025.14.5.343
Self-Supervised Learning Waste Classification Reconstruction Learning contrastive learning representation learning
An Android Malware Detection Technique Using Gaussian Mixture Model Clustering
Seung Min Lee  Seok Hyun Ahn  Seong-Je Cho  Dong Jae Kim  Young Sup Hwang
Machine learning-based techniques have been extensively explored for detecting malicious Android applications. However, traditional models often suffer from performance degradation over time due to concept drift, where the behavioral and structural features of apps evolve. To address this issue, we propose a novel detection framework that leverages Gaussian Mixture Model (GMM) clustering to mitigate the impact of concept drift. Our approach models the underlying data distribution as a mixture of Gaussian components and trains a specialized classifier for each component. This allows the system to adapt to shifting feature distributions without the need for frequent retraining. Experimental evaluations conducted on Android app datasets spanning from 2014 to 2023 demonstrate that traditional machine learning models experience significant performance decline on post-2019 data due to concept drift. In contrast, our GMM-based framework maintains robust detection performance across all years, achieving an 8.9 percentage point improvement in F1-score and a 10.1 percentage point increase in Area Under Time (AUT) compared to conventional methods.
The Transactions of the Korea Information Processing Society, Vol. 14, No. 5, pp. 352-362, May. 2025
https://doi.org/10.3745/TKIPS.2025.14.5.352
Android Malware Concept Drift Gaussian mixture model clustering Area under Time
Evaluating Hallucination in LLM with Jeju Dialect Inputs
Seokjae Gwon  Jeongho Lee  Gyeonghoe Kim  Heeyeong Suh  Seyoung Lee
LLM have been actively utilized in various high-stakes applications. However, as most models are trained on standard language data, they exhibit structural limitations in processing low-resource linguistic inputs such as dialects. In particular, when faced with inputs in the Jeju dialect—which significantly differs from Standard Korean in grammar, vocabulary, and expression—LLMs often fail to accurately interpret user intent and generate hallucinated content without factual basis. This study empirically investigates this issue by constructing a dataset of queries in both Standard Korean and the Jeju dialect. A quantitative comparison was conducted on commercial LLM focusing on response reliability, hallucination frequency, sensitivity to linguistic components, and model-specific response divergence. Experimental results show that dialectal input significantly increases the error rate, with hallucination being more prominent in multiple-answer formats and queries with complex verb endings. This study provides concrete evidence of LLM's heightened sensitivity to dialectal inputs and offers foundational insights for developing dialect-aware language models and designing safer generative AI systems
The Transactions of the Korea Information Processing Society, Vol. 14, No. 5, pp. 363-371, May. 2025
https://doi.org/10.3745/TKIPS.2025.14.5.363
Hallucination Large Language Model Dialect Linguistic Diversity
Ppormer: A PPR-based Diffusion Transformer for Effective Node Representation Learning on Weighted Graphs
Park Siyeon  Lee Ki Yong
This paper proposes Ppormer (PPR-based Diffusion Transformer), a novel graph neural network designed to improve node representation learning on weighted graphs. Ppormer incorporates three complementary types of information into the node representation process: (1) Global contextual information, captured via diffusion-based self-attention mechanisms that model semantic relationships across the entire graph; (2) Local structural information, obtained through Graph Convolutional Network (GCN)-based message passing over immediate neighbors; and (3) Global structural information, derived from Personalized PageRank (PPR) to reflect long-range topological relevance based on edge strength. These heterogeneous signals are adaptively fused using the proposed FusionAttention module. Experiments conducted on the Cora dataset, with edge weights computed using Jaccard, Canberra, and Euclidean similarities, demonstrate that Ppormer achieves accuracy scores of 84.00%, 83.94%, and 85.54%, respectively—outperforming all baseline models. These results validate the effectiveness and generalizability of Ppormer in various weighted graph scenarios.
The Transactions of the Korea Information Processing Society, Vol. 14, No. 5, pp. 372-378, May. 2025
https://doi.org/10.3745/TKIPS.2025.14.5.372
Graph Neural Networks Weighted Graphs Node Representation Diffusion Transformer Attention Mechanism