Identifying Unusual Days

Kim, Min-Kyong,Kotz, David Korean Institute of Information Scientists and Eng 2011 Journal of Computing Science and Engineering Vol.5 No.1

Pervasive applications such as digital memories or patient monitors collect a vast amount of data. One key challenge in these systems is how to extract interesting or unusual information. Because users cannot anticipate their future interests in the data when the data is stored, it is hard to provide appropriate indexes. As location-tracking technologies, such as global positioning system, have become ubiquitous, digital cameras or other pervasive systems record location information along with the data. In this paper, we present an automatic approach to identify unusual data using location information. Given the location information, our system identifies unusual days, that is, days with unusual mobility patterns. We evaluated our detection system using a real wireless trace, collected at wireless access points, and demonstrated its capabilities. Using our system, we were able to identify days when mobility patterns changed and differentiate days when a user followed a regular pattern from the rest. We also discovered general mobility characteristics. For example, most users had one or more repeating mobility patterns, and repeating mobility patterns did not depend on certain days of the week, except that weekends were different from weekdays.

Application of Improved Variational Recurrent Auto-Encoder for Korean Sentence Generation

Sangchul Hahn(한상철),Seokjin Hong(홍석진),Heeyoul Choi(최희열) Korean Institute of Information Scientists and Eng 2018 정보과학회논문지 Vol.45 No.2

Due to the revolutionary advances in deep learning, performance of pattern recognition has increased significantly in many applications like speech recognition and image recognition, and some systems outperform human-level intelligence in specific domains. Unlike pattern recognition, in this paper, we focus on generating Korean sentences based on a few Korean sentences. We apply variational recurrent auto-encoder (VRAE) and modify the model considering some characteristics of Korean sentences. To reduce the number of words in the model, we apply a word spacing model. Also, there are many Korean sentences which have the same meaning but different word order, even without subjects or objects; therefore we change the unidirectional encoder of VRAE into a bidirectional encoder. In addition, we apply an interpolation method on the encoded vectors from the given sentences, so that we can generate new sentences which are similar to the given sentences. In experiments, we confirm that our proposed method generates better sentences which are semantically more similar to the given sentences.

Designing Technology for Visualisation of Interactions on Mobile Devices

Deray, Kristine,Simoff, Simeon Korean Institute of Information Scientists and Eng 2009 Journal of Computing Science and Engineering Vol.3 No.4

Interactions are intrinsic part of what we do. We interact when we work, when we learn, when we visit a doctor, and when we play. With the advent of information and communications technology we can collect rich data (video, audio, and various transcripts including text chat) about such interactions. This opens an opportunity to monitor the dynamics of interactions and to get deeper insights of how they unfold and deliver this information to the interacting parties. This paper presents the design of a technology for visualising information about the dynamics of unfolding of interactions and presenting it in an ambient display on mobile devices. The purpose of this technology is the delivery of such information to the point of decision making.

Design and Development of a Multimodal Biomedical Information Retrieval System

Demner-Fushman, Dina,Antani, Sameer,Simpson, Matthew,Thoma, George R. Korean Institute of Information Scientists and Eng 2012 Journal of Computing Science and Engineering Vol.6 No.2

The search for relevant and actionable information is a key to achieving clinical and research goals in biomedicine. Biomedical information exists in different forms: as text and illustrations in journal articles and other documents, in images stored in databases, and as patients' cases in electronic health records. This paper presents ways to move beyond conventional text-based searching of these resources, by combining text and visual features in search queries and document representation. A combination of techniques and tools from the fields of natural language processing, information retrieval, and content-based image retrieval allows the development of building blocks for advanced information services. Such services enable searching by textual as well as visual queries, and retrieving documents enriched by relevant images, charts, and other illustrations from the journal literature, patient records and image databases.

Automatic Product Review Helpfulness Estimation based on Review Information Types

Munhyong Kim(김문형),Hyopil Shin(신효필) Korean Institute of Information Scientists and Eng 2016 정보과학회논문지 Vol.43 No.9

Many available online product reviews for any given product makes it difficult for a consumer to locate the helpful reviews. The purpose of this study was to investigate automatic helpfulness evaluation of online product reviews according to review information types based on the target of information. The underlying assumption was that consumers find reviews containing specific information related to the product itself or the reliability of reviewers more helpful than peripheral information, such as shipping or customer service. Therefore, each sentence was categorized by given information types, which reduced the semantic space of review sentences. Subsequently, we extracted specific information from sentences by using a topic-based representation of the sentences and a clustering algorithm. Review ranking experiments indicated more effective results than other comparable approaches.

Improvement in Network Intrusion Detection based on LSTM and Feature Embedding

Hyeokmin Gwon(권혁민),Chungjun Lee(이청준),Rakun Keum(금락운),Heeyoul Choi(최희열) Korean Institute of Information Scientists and Eng 2021 정보과학회논문지 Vol.48 No.4

Network Intrusion Detection System (NIDS) is an essential tool for network perimeter security. NIDS inspects network traffic packets to detect network intrusions. Most of the existing works have used machine learning techniques for building the system. While the reported works demonstrated the effectiveness of various artificial intelligence algorithms, only a few of them have utilized the time-series information of network traffic data. Also, categorical information of network traffic data has not been included in neural network-based approaches. In this paper, we propose network intrusion detection models based on sequential information using the long short-term memory (LSTM) network and categorical information using the embedding technique. We have conducted experiments using models with UNSW-NB15, which is a comprehensive network traffic dataset. The experiment results confirm that the proposed method improves the performance, with a binary classification accuracy rate of 99.72%.

Protecting the iTrust Information Retrieval Network against Malicious Attacks

Chuang, Yung-Ting,Melliar-Smith, P. Michael,Moser, Louise E.,Lombera, Isai Michel Korean Institute of Information Scientists and Eng 2012 Journal of Computing Science and Engineering Vol.6 No.3

This paper presents novel statistical algorithms for protecting the iTrust information retrieval network against malicious attacks. In iTrust, metadata describing documents, and requests containing keywords, are randomly distributed to multiple participating nodes. The nodes that receive the requests try to match the keywords in the requests with the metadata they hold. If a node finds a match, the matching node returns the URL of the associated information to the requesting node. The requesting node then uses the URL to retrieve the information from the source node. The novel detection algorithm determines empirically the probabilities of the specific number of matches based on the number of responses that the requesting node receives. It also calculates the analytical probabilities of the specific numbers of matches. It compares the observed and the analytical probabilities to estimate the proportion of subverted or non-operational nodes in the iTrust network using a window-based method and the chi-squared statistic. If the detection algorithm determines that some of the nodes in the iTrust network are subverted or non-operational, then the novel defensive adaptation algorithm increases the number of nodes to which the requests are distributed to maintain the same probability of a match when some of the nodes are subverted or non-operational as compared to when all of the nodes are operational. Experimental results substantiate the effectiveness of the detection and defensive adaptation algorithms for protecting the iTrust information retrieval network against malicious attacks.

INFINITE: Intelligent Neural Foley Inference Network for Inter-scene Temporal Events

Seongmin Lee(이성민),Yewon Jeong(정예원),Jeongseon Yu(유정선),Donghoon Cha(차동훈),Jaewon Moon(문재원),Hanwool Sul(설한울) Korean Institute of Information Scientists and Eng 2025 한국정보과학회 학술발표논문집 Vol.2025 No.7

In modern content production pipelines, Foley sound design remains one of the most labor-intensive post-production tasks[1]. Existing Video-to-Audio generation approaches show promise in producing natural sounds aligned with video content but often fail in the context of sound effects where semantic and temporal alignment is critical, especially during scene transitions. In this paper, we introduce INFINITE (Intelligent Neural Foley Inference Network for Inter-scene Temporal Events), a Video-to-Foley model that generates Foley sound robustly even in the presence of scene transitions. Unlike prior models[2, 3, 4], which average frame-wise embeddings and thus fail to adapt to shifting semantics, our method extracts semantic embeddings at each detected action onset and conditions the generator accordingly. This ensures robust semantic alignment even when video context shifts dramatically. Our method significantly outperforms the baseline in terms of Frechet Audio Distance (FAD).

Attention-Guided Pseudo Caption Dynamic Filtering for Factual Figure Captioning

Taehwan Kim(김태환),Jieun Woo(우지은),Seongwan Park(박성완),Yun-Gyung Cheong(정윤경) Korean Institute of Information Scientists and Eng 2024 한국정보과학회 학술발표논문집 Vol.2024 No.12

Malware Detection and Prevention in Internet of Things Networks Using Deep Learning

Quyen Le Thi Nguyen,Duc-Tai Le,Hyunseung Choo Korean Institute of Information Scientists and Eng 2025 한국정보과학회 학술발표논문집 Vol.2025 No.7

This paper presents a deep learning-based approach for detecting malware in IoT networks, focusing on malicious traffic generated by specific bots and malware. The proposed system addresses key IoT security challenges, including evolving threats and device resource limitations, by leveraging advanced neural network architectures. Our framework aims to enhance detection accuracy and scalability while maintaining data integrity and confidentiality across interconnected IoT environments. To this end, the study explores the integration of hybrid neural models with privacy-preserving mechanisms to ensure efficient and secure deployment in real-world IoT scenarios.