• Transactions on Semiconductor Engineering
• /
• v.2 no.1
• /
• pp.17-20
• /
• 2024

This paper presents a novel lightweight object detection model tailored for low-powered edge devices, addressing the limitations of traditional resource-intensive computer vision models. Our proposed detector, inspired by the Single Shot Detector (SSD), employs a compact yet robust network design. Crucially, it integrates an 'enhancer block' that significantly boosts its efficiency in detecting smaller objects. The model comprises two primary components: the Light_Block for efficient feature extraction using Depth-wise and Pointwise Convolution layers, and the Enhancer_Block for enhanced detection of tiny objects. Trained from scratch on the Udacity Annotated Dataset with image dimensions of 300x480, our model eschews the need for pre-trained classification weights. Weighing only 5.5MB with approximately 0.43M parameters, our detector achieved a mean average precision (mAP) of 27.7% and processed at 140 FPS, outperforming conventional models in both precision and efficiency. This research underscores the potential of lightweight designs in advancing object detection for edge devices without compromising accuracy.

• The Journal of Korea Institute of Information, Electronics, and Communication Technology
• /
• v.16 no.6
• /
• pp.540-546
• /
• 2023

The scope of IoT use has expanded due to the development of related technologies, and various sensors have been developed and distributed to meet the demand for implementing various services. Measuring alcohol concentration using a sensor can be used to prevent drunk driving, and to make this possible, accurate alcohol concentration must be measured and safe transmission from the smartphone to the server must be guaranteed. Additionally, a process of converting the measured alcohol concentration value into a standard value for determining the level of drinking is necessary. In this paper, we propose and implement a system. Security with remote servers applies SSL at the network layer to ensure data integrity and confidentiality, and the server encrypts the received information and stores it in the database to provide additional security. As a result of analyzing the accuracy of alcohol concentration measurement and communication efficiency, it was confirmed that the measurement and transmission were within the error tolerance.

• The Journal of the Korea institute of electronic communication sciences
• /
• v.19 no.1
• /
• pp.189-196
• /
• 2024

Frequent false positives alarm from the Intelligent Selective Control System have raised significant concerns. These persistent issues have led to declines in operational efficiency and market credibility among agents. Developing a new model or replacing the existing one to mitigate false positives alarm entails substantial opportunity costs; hence, improving the quality of the training dataset is pragmatic. However, smaller organizations face challenges with inadequate capabilities in dataset collection and refinement. This paper proposes an automatic human pose data collection system centered around a human pose estimation model, utilizing camera-based sensor fusion techniques and edge devices. The system facilitates the direct collection and real-time processing of field data at the network periphery, distributing the computational load that typically centralizes. Additionally, by directly labeling field data, it aids in constructing new training datasets.

• Convergence Security Journal
• /
• v.23 no.4
• /
• pp.93-100
• /
• 2023

This paper presents a novel approach for assigning group keys within a dynamic swarm unmanned system environment. In this environment, multiple groups of unmanned systems have the flexibility to merge into a single group or a single unmanned system group can be subdivided into multiple groups. The proposed protocol encompasses two key steps: group key generation and sharing. The responsibility of generating the group key rests solely with the leader node of the group. The group's leader node employs a secret sharing scheme to fragment the group key into multiple fragments, which are subsequently transmitted. Nodes that receive these fragments reconstruct a fresh group key by combining their self-generated secret fragment with the fragment obtained from the leader node. Subsequently, they validate the integrity of the derived group key by employing the hash function. The efficacy of the proposed technique is ascertained through an exhaustive assessment of its security and communication efficiency. This analysis affirms its potential for robust application in forthcoming swarm unmanned system operations scenarios characterized by frequent network group modifications.

• The Transactions of the Korea Information Processing Society
• /
• v.13 no.3
• /
• pp.122-129
• /
• 2024

Though programming education has been stressed recently, the elementary, middle, and high school students are having trouble in programming education. Most programming environments for them are based on block coding, which hinders them from moving to text coding. The traditional PC environment has also troubles such as maintenance problems. In this situation, mobile applications can be considered as alternative programming environments. This paper addresses the design and implementation of coding applications for mobile devices. As a prototype, a Scheme interpreter mobile app is proposed, where Scheme is used for programming courses at MIT since it supports multi-paradigm programming. The implementation has the advantage of not consuming the network bandwidth since it is designed as a standalone application. According to the benchmark result, the execution time on Android devices, relative to that on a desktop, was 131% for the Derivative and 157% for the Tak. Further, the maximum execution times for the benchmark programs on the Android device were 19.8ms for the Derivative and 131.15ms for the Tak benchmark. This confirms that when selecting an Android device for programming education purposes, there are no significant constraints for training.

• Convergence Security Journal
• /
• v.24 no.3
• /
• pp.153-163
• /
• 2024

Despite the increasing hacking threats and security threats as IT technology advances and many companies adopt security solutions, cyberattacks and threats still persist for years. APT attack is a technique of selecting a specific target and continuing to attack. The threat of an APT attack uses all possible means through the electronic network to perform APT for years. Zero-day attacks, malicious code distribution, and social engineering techniques are performed, and some of them directly invade companies. These techniques have been in effect since 2000, and are similarly used in voice phishing, especially for social engineering techniques. Therefore, it is necessary to study countermeasures against APT attacks. This study analyzes the attack cases of Korean-based APT groups in Korea and suggests the correct method of using intelligence to analyze APT attack groups.

• The Transactions of the Korea Information Processing Society
• /
• v.13 no.8
• /
• pp.367-373
• /
• 2024

The growth of mobile devices and network infrastructure has brought significant changes to the music industry. Online streaming services has allowed music consumption without constraints of time and space, leading to increased consumer engagement in music creation and sharing activities, resulting in a vast accumulation of music data. In this study, we define metadata as "song sentences" by using a user's playlist. To calculate similarity, we embedded them into a high-dimensional vector space using skip-gram with negative sampling algorithm. Performance eva luation results indicated that the recommended music algorithm, utilizing singers, genres, composers, lyricists, arrangers, eras, seasons, emotions, and tag lists, exhibited the highest performance. Unlike conventional recommendation methods based on users' behavioral data, our approach relies on the inherent information of the tracks themselves, potentially addressing the cold start problem and minimizing filter bubble phenomena, thus providing a more convenient music listening experience.

• Convergence Security Journal
• /
• v.21 no.4
• /
• pp.59-68
• /
• 2021

The Future Battlefield includes the main areas of modern warfare, including the ground, sea, and air, as well as cyberspace and space. Cyberspace consists of computers, wired and wireless networks, and spans the ground, sea, air, and space domains. Cyber warfare takes place in cyberspace, so it is not easy for people without expertise in cyber to recognize the cyber situation. Therefore, training personnel with professional knowledge and skills in cyber is paramount in preparation for cyber warfare. In particular, the results of cyber warfare will vary greatly depending on the ability of cyber combatants to carry it out, the performance of cyber systems, and the proficiency of cyber warfare procedures. The South Korean military has power to respond to cyber warfare at various levels, centering on the Cyber Operations Command, but there is a limit to defending all the rapidly expanding cyberspace. In this paper, to overcome these limitations, we looked at the changes in Germany's cyber warfare response policy. Based on them, the organization structure, weapon system, and education and training system of future Korean military cyber forces are presented separately.

• Journal of Internet of Things and Convergence
• /
• v.10 no.5
• /
• pp.13-18
• /
• 2024

The application of machine learning models for trend prediction in the domestic stock market is increasing. In particular, utilizing machine learning is essential for analyzing and predicting complex time-series data, such as stock price data. This study proposes a machine learning system for financial time-series trend prediction, utilizing cloud computing services. First, for data collection, the serverless service of Amazon Web Services was employed, and the thresholds of technical analysis indicators were optimized through a genetic algorithm. The optimized indicators were then used as training data for Echo State Network, Recurrent Neural Network (RNN), and various machine learning classification models to predict the trend of each stock. Based on the predicted trends, backtesting was conducted, and the results showed that the average returns were 334% for ESN, 175% for RNN, and 199% for classification models. Therefore, this study suggests that machine learning exhibits high predictive power in domestic stock investment and holds various potential applications.

• Journal of the Korean Geographical Society
• /
• v.29 no.4
• /
• pp.402-419
• /
• 1994

Telecommunication and intormation technology have been conceived as crucial as well as revolutionary elements for recent and future social and economic development, and their development have led to a spatial reorganization and locational change of economic activities. Information technology has resulted in important changes in the organization structure and location of firm. This study draws attention to the understanding of the relationship between the diffusion of information technology and changes in firm activities with the special reference to the service industry of the United States. Information technology has had a significant impact on the growth and changes of the service industry of the United States through changes in the organizational and employment structure, market structure, and locational changes. The impact of information technology on location changes of the service industry shows two opposite patterns, concentration and decentralization. Among these patterns, the location change in the service industry of the United States reveals predominantly the decentralization tendency such as suburbanization and transfer to lower ranking cities rather than concentration. In case of Korea, however, it is anticipated that the rapid development of information technology may lead to the concentration of the service industry in Seoul and Capital region.