Il Kon Kim

Kernel based intrusion detection system

Recently, applying artificial intelligence, machine learning and data mining techniques to intrusion detection system are increasing. But most of researches are focused on improving the performance of classifier. Selecting important features from input data lead to a simplification of the problem, faster and more accurate detection rates. Thus, selecting important features is an important issue in intrusion detection. Another issue in intrusion detection is that most of the intrusion detection systems are performed by off-line and it is not proper method for realtime intrusion detection system. In this paper, we develop the realtime intrusion detection system which combines on-line feature extraction method with least squares support vector machine classifier. Applying proposed system to KDD CUP 99 data, experimental results show that it has remarkable performance compared to off-line intrusion detection system.

Adoption of a PDA-based home hospice care system for cancer patients.

This study describes a systematic and efficient point-of-care home hospice information system, which can effectively manage patient information and can easily be incorporated into a home hospice service using wired and/or wireless communications and mobile computing technology. To help with the design of a portable home hospice information system, we analyzed the services provided by visiting nurses and the opinions of users, based on the 2005 Guidelines for Cancer Patient Management Program of the Ministry of Health and Welfare of Korea. We performed a service trial with three experienced visiting nurses. A visiting nurse can use a PDA, implementing the hospice information system, to accomplish tasks that range from registering a new patient to performing nursing services based on the patients condition, as well as accessing guidelines and searching past information for better nursing performance. The system resulted in a saving of 8.5 minutes in recoding time. We expect the findings of this study to help field workers in community nursing care by reducing unnecessary paperwork and by enabling improvements in the quality of point-of-care services.

Incremental Feature Extraction Based on Empirical Kernel Map

A new incremental kernel principal component analysis is proposed for the nonlinear feature extraction from the data. The problem of batch kernel principal component analysis is that the computation becomes prohibitive when the data set is large . Another problem is that, in order to update the eigenvectors with another data, the whole decomposition from scratch should be recomputed. The proposed method overcomes these problems by incrementally update eigenspace and using empirical kernel map as kernel function. The proposed method is more efficient in memory requirement than a batch kernel principal component and can be easily improved by re-learning the data. In our experiments we show that proposed method is comparable in performance to a batch kernel principal component for the classification problem on nonlinear data set.

SOA-based integrated pervasive personal health management system using PHDs

Nowadays, the need for effective integrated health information management is very high, with which each individual can monitor his/her health status and the ever increasing healthcare cost can be significantly reduced therefore. Moreover, advancements in information, communication and sensor technology have made readily available many types of highly accurate yet affordable PHDs (Personal Health Devices) including glucose meter and blood pressure monitor just to name a few. In this paper, we present our experience with the design and implementation of an integrated pervasive personal health management system with the following features: (a) PHDs with wireless communication capabilities so that various kinds of health data can be pervasively captured; (b) standard interfaces so that various components can interacts with each other in interoperable manners; and (c) a SOA (Service Oriented Architecture)-based service architecture so that customized services can be easily implemented using well-defined reusable components.

Integrating Clinical Information in National Biobank of Korea

The National Biobank of Korea (NBK) is a government supported project that aims at consolidating various human-originated biomedical resources collected by individual hospitals nation-wide and integrating them with their donors’ clinical information which researchers can take advantage of. In this paper, we present our experiences in developing the Clinical Information Integration System (CIIS) for NBK. The system automatically extracts clinical data from hospital information systems as much as possible to avoid errors from manual entry by human errors. It maintains the independence of individual hospitals by employing a two-layer approach, one of which takes care of all hospital-specific aspects. Interoperability is achieved by adopting HL7 v2.x messaging between the biobank and hospitals. We report the current status of the biobank and system deployments. We finally identify limitations and discuss how to improve them.

Processing HL7-CDA Entry for Semantic Interoperability

Semantic interoperability is currently one of the most challenging topics in health information area. Context agreement between physicians is needed in case that a patient has to consult a variety of physicians for the treatment of one medical problem. To support semantic interoperability, identical syntactic representation is also considered. This study solves the semantic interoperability problem based on health information standard. HL7 v3 and CDA entry is well known for the formal health information model that defines the conceptual classes to map healthcare environment and resources. CDA entry provides the basis of the formalization of medical information related clinical statement. This study proposes the method of processing CDA entry for semantic interoperability, and describes how to analysis entries and extract the data items.

CDA Generation and Integration for Health Information Exchange Based on Cloud Computing System

Successful deployment of Electronic Health Record helps improve patient safety and quality of care, but it has the prerequisite of interoperability between Health Information Exchange at different hospitals. The Clinical Document Architecture (CDA) developed by HL7 is a core document standard to ensure such interoperability, and propagation of this document format is critical for interoperability. Unfortunately, hospitals are reluctant to adopt interoperable HIS due to its deployment cost except for in a handful countries. A problem arises even when more hospitals start using the CDA document format because the data scattered in different documents are hard to manage. In this paper, we describe our CDA document generation and integration Open API service based on cloud computing, through which hospitals are enabled to conveniently generate CDA documents without having to purchase proprietary software. Our CDA document integration system integrates multiple CDA documents per patient into a single CDA document and physicians and patients can browse the clinical data in chronological order. Our system of CDA document generation and integration is based on cloud computing and the service is offered in Open API. Developers using different platforms thus can use our system to enhance interoperability.

Health innovation for aging society.

The trend of increasing population of the elderly is occurring not only in Japan and other developed countries but also in nations across the globe. While this trend speaks well for ongoing improvements in health, welfare, diet and other factors, the attendant demographics will also result in many challenges with regard to managing the societal impact, including management of increasing costs of care required as a population ages. In order to adapt to aged societies, and to support a greater desire of people to live independently as well as to enable younger generation to maintain their health and active participation in the society when they become aged in the future, we need a significant change in the behaviors and thought processes of healthcare providers, the public, and elderly peoples friends and families. This will relate to all aspects of a persons life, including planning, decision-making, and day to day living. To make super-aged societies sustainable, drastic innovations related to the social system are necessary to enhance peoples happiness and reduce social costs at the same time [1].

Machine learning approach to realtime intrusion detection system

Computer security has become a critical issue with the rapid development of business and other transaction systems over the internet. Recently applying artificial intelligence, machine learning and data mining techniques to intrusion detection system are increasing. But most of researches are focused on improving the classification performance of classifier. Selecting important features from input data lead to a simplification of the problem, faster and more accurate detection rates. Thus selecting important features is an important issue in intrusion detection. Another issue in intrusion detection is that most of the intrusion detection systems are performed by off-line and it is not proper method for realtime intrusion detection system. In this paper, we develop the realtime intrusion detection system which combining on-line feature extraction method with Least Squares Support Vector Machine classifier. Applying proposed system to KDD CUP 99 data, experimental results show that it have remarkable feature feature extraction and classification performance compared to existing off-line intrusion detection system.

Feature Extraction and Classification System for Nonlinear and Online Data

A novel incremental feature extraction and classification system is proposed. Kernel PCA is famous nonlinear feature extraction method. The problem of Kernel PCA is that the computation becomes prohibitive when the data set is large. Another problem is that, in order to update the eigenvectors with another data, the whole eigenspace should be recomputed. Proposed feature extraction method overcomes these problems by incrementally eigenspace update and using empirical kernel map as kernel function. Proposed feature extraction method is more efficient in memory requirement than a Kernel PCA and can be easily improved by re-learning the data. For classification extracted features are used as input for Least Squares SVM. In our experiments we show that proposed feature extraction method is comparable in performance to a Kernel PCA and proposed classification system shows a high classification performance on UCI benchmarking data and NIST handwritten data set.