연합학습에서의 손실함수의 적응적 선택을 통한 효과적인 적대적 학습

이수철 ( Suchul Lee ) 한국인터넷정보학회 2024 인터넷정보학회논문지 Vol.25 No.2

Although federated learning is designed to be safer than centralized methods in terms of security and privacy, it still has many vulnerabilities. An attacker performing an adversarial attack intentionally manipulates the deep learning model by injecting carefully crafted input data, that is, adversarial examples, into the client's training data to induce misclassification. A common defense strategy against this is so-called adversarial training, which involves preemptively learning the characteristics of adversarial examples into the model. Existing research assumes a scenario where all clients are under adversarial attack, but considering the number of clients in federated learning is very large, this is far from reality. In this paper, we experimentally examine aspects of adversarial training in a scenario where some of the clients are under attack. Through experiments, we found that there is a trade-off relationship in which the classification accuracy for normal samples decreases as the classification accuracy for adversarial examples increases. In order to effectively utilize this trade-off relationship, we present a method to perform adversarial training by adaptively selecting a loss function depending on whether the client is attacked.

연구개발 프로젝트 정성·정량평가 비교 분석을 통한 성과평가 발전방향 연구

이수철(Suchul Lee),고미현(Mihyun Ko) 한국기술혁신학회 2017 기술혁신학회지 Vol.20 No.2

본 연구에서는 정부출연연구기관 중 하나인 K 연구원 사례를 활용하여 연구개발 프로젝트에 대한 정성평가 결과와 정량평가 결과를 비교 분석함으로써 연구개발 프로젝트 성과평가 체계 개선을 위한 시사점을 도출하였다. 비교 분석을 위해 정성평가 결과는 국가과학기술연구회 규정에 따라 진행된 실제 평가결과를, 정량평가 결과는 연구개발 프로젝트의 산출물 데이터를 기반으로 자료포락분석(data envelopment analysis) 결과를 활용하였다. 비모수 상관관계 분석(Kendall’s τ)을 비롯한 다양한 분석결과, 정성평가와 정량평가 결과 사이에는 분명한 차이점이 있는 것으로 확인되었다. 따라서 단순 정성 또는 정량평가로 연구개발 프로젝트를 평가하기보다 정성 및 정량평가가 상호보완적으로 진행된다면, 평가 시 각 연구개발 프로젝트의 특성을 반영함과 동시에 객관성도 보다 확보할 수 있을 것으로 기대된다. This study measures and compares the performance of research and development (R&D) programs in government-funded research institutes (GRIs) in terms of qualitative and quantitative approaches to find out strategic insights for improving performance evaluation policy. In particular, we adopt the evaluation results from the real data of K institute in 2015 for a qualitative evaluation and the results of data envelopment analysis (DEA) for a quantitative evaluation. Comparative analysis of the R&D performance of 14 programs finds that the difference between the evaluation results of qualitative and quantitative approaches is significant. From this finding, we suggest several strategic directions to complement two approaches each other.

슈퍼컴퓨터 활용 계산과학 기반 연구개발 효율화 경제성 분석

이수철(Suchul Lee),이용호(Yongho Lee) 한국경영과학회 2016 한국경영과학회 학술대회논문집 Vol.2016 No.10

This paper aims to analyze an economic feasibility for improving R&D efficiency using computational science with supercomputers employing Cost-benefit analysis. In particular, we analyzed time-saving effect based on past applying data by 17 science and technology categories, and then estimated future economic values assuming that computational science with supercomputers was applied to large-scale national R&D projects at several government-funded research institutes for next five years. In this study, we confirms it is feasible enough that applying computational science to national R&D projects under various situations because the B/C ratios are over one.

중소기업 제조 공정 혁신 기술 도입 활성화를 위한 디지털 트윈 기술 적용 - 협동 로봇 적용 사례 -

이수철(Suchul Lee),최상수(SangSu Choi),유태종(Taejong Yoo) (사)한국CDE학회 2022 한국CDE학회 논문집 Vol.27 No.4

This study is to provide a simulation environment by applying only the minimum technology based on digital twin and to provide an easier decision-making environment for SMEs to introduce the innovative manufacturing process. A framework that utilizes a combination of resource library, layout design, and simulation algorithm based on digital twin technology was proposed. In order to verify the applicability and limitations of the proposed methodology, a comparison of productivity performance indicators was performed when the manual process part was replaced with a collaborative robot in the high frequency heat treatment process for saw blade manufacturing.

Cumulative DEA/Malmquist Index 기법을 이용한 정부출연 연구기관 연구개발 효율성 변화 분석

이수철(Suchul Lee),이동호(Dong Ho Lee) 한국경영과학회 2016 한국경영과학회지 Vol.41 No.1

This paper presents a framework to analyze the change in the research and development (R&D) efficiency of government-funded research institutes (GRIs) in Korea. Cumulative data envelopment analysis/Malmquist index method is utilized to analyze the changes in R&D efficiency of GRIs. Data analysis of the R&D activities of 10 GRIs in Korea Research Council of Fundamental Science & Technology showed that the average R&D efficiency of the 10 GRIs improved from 2009 to 2013. However, the efficiency of a few GRIs decreased in terms of the catch-up index. The proposed framework can help management teams diagnose the current state of R&D activities and determine the efficacy of strategic actions by comparing efficiencies in the past.

D-Fi: 다이버시티를 이용하기 위한 OFDM기반 Bloom Filter 무선랜 프로토콜

이수철(Suchul Lee),임상순(Sangsoon Lim),이준락(Jun-Rak Lee),김종권(Chong-Kwon Kim) 한국정보과학회 2013 정보과학회논문지 : 정보통신 Vol.40 No.4

다이버시티를 이용하기 위해서는 변화하는 무선채널의 상태를 측정해야 한다. 하지만 채널 상태 측정은 데이터전송에 사용되어야 하는 시간/주파수 자원을 사용해야 하므로, 정확한 채널상태를 얻는 것은 프로토콜 측면에서 오버헤드를 야기한다. 본 논문에서는, 프로토콜 효율성을 유지한 채로 채널상태를 측정할 수 있는 무선랜 프로토콜인 D-Fi를 제안한다. 제안하는 프로토콜 D-Fi는 채널경쟁을 하는 동안 OFDM기반 블룸필터를 이용하여 채널상태를 측정한다. 블룸필터가 가지는 내재적 약점을 해결하기 위해서, 분석기반 멀티채널 백오프 알고리즘을 제안하고 나아가서 기계학습을 블룸필터에 적용하였다. 우리는 D-Fi 프로토콜을 USRP/GNURadio 플랫폼에 실제로 구현함으로써, 제안하는 프로토콜이 매우 실용적이며, 다이버시티를 효율적으로 이용함을 보였다. To exploit frequency diversity in Wi-Fi channels, instantaneous channel quality must be estimated. However, there is a trade-off between acquiring channel quality information and improving protocol efficiency because channel estimation consumes time and frequency resource that ideally should be used for data transfer. In this paper, we present D-Fi (Diversity-aware Wi-Fi), a novel Wi-Fi PHY/MAC protocol, that capitalizes on frequency diversity gains while sustaining protocol efficiency. The D-Fi design allows to estimate channel quality while D-Fi is performing channel contention using an OFDM-based Bloom filter. To resolve the ambiguity caused by the Bloom filter, we adopt two methods: (i) An analysis-based multi-channel backoff method enables to explore/exploit frequency diversity while reducing the occurrence of the ambiguity. (ii) Applying machine learning (ML) methods to the D-Fi PHY/MAC protocol corrects the ambiguity taken place already and makes our protocol reliable. We have shown the feasibility of D-Fi by implementing it on the USRP/GNURadio platform. Experiments and trace-driven simulations show that D-Fi successfully achieves frequency.

정부출연 연구기관 경쟁력 분석 모형 개발

이수철 ( Lee Suchul ),서의호 ( Suh Euiho ),이재영 ( Lee Jaeyoung ) 경남대학교 산업경영연구소 2013 지역산업연구 Vol.36 No.2

Korean government has established the Government Funded Research Institutes (GFRIs) for achieving national competitiveness in science, economy, and society areas since 1966. Recently, Korean government pay increasing attention to evaluating GFRIs’ performance for improving their efficiency of research performance and research competitiveness, mainly because the R&D ability of industry and university has been increased dramatically. The purpose of this paper is to suggest the competitiveness analysis model for GFRI. To validate the usefulness of the proposed model, the case study was conducted with the 13 GFRIs included in Korea Research Council of Fundamental Science & Technology (KRCF). In particular, the model adopt four perspectives in Porter’s Diamond model (Porter, 1990) to analyze the competitiveness of KRCF GFRIs from the various perspectives.

미래 인터넷 테스트베드 모니터링 시스템의 구현

이수철(Suchul Lee),박근모(Keunmo Park),최환석(Hwanseok Choi),김종권(Chong-Kwon Kim) 한국정보과학회 2011 정보과학회논문지 : 정보통신 Vol.38 No.5

급격한 성공과 더불어 현재의 인터넷은 사용자의 규모, 대상적 변화로 인하여 보안, 성능 등등의 각종 문제에 봉착하였다. 따라서 근본적인 인터넷의 문제를 해결하기 위하여 세계 연구 계에서는 혁신적 ‘미래 인터넷 연구’가 진행되고 있다. 혁신적 미래 인터넷 기술 연구에서는 새롭게 제안된 혁신적인 기술들을 시험 및 검증하기 위한 실험용 테스트 베드 구축이 미래인터넷 연구의 선행과제로 무엇보다 중요하다. 본고에서는 미래 인터넷 테스트 베드 관리 및 운영 시스템인 dvNOC(Distributed Virtual Network Operations Center)을 제안한다. 또한 dvNOC의 디자인 및 구현내용을 소개한다. dvNOC은 국내 최초 테스트 베드 망 관리 시스템으로써, 미국의 선도 그룹인 인디애나 대학의 망 관리 시스템인 GMOC과의 페더레이션되어 전세계의 테스트 베드를 모니터링 한다. Increasingly popular Internet has faced several problems such as performance degradation, security related problems, etc. To solve the fundamental problem of the internet, ‘Future Internet” community has endeavored to study revolutionary future internet architectures ceaselessly. Among these future internet researches, proposing world-wide future internet testbed is a very important issue since they could be used to experiment and verify the newly proposed future internet schemes. In this paper, we propose dvNOC (Distributed Virtual Network Operations Center), a management system for monitoring testbed’s resources. We design and implement dvNOC. To the best of our knowledge, dvNOC is the first (in Korea) system which monitors and maintains resources of the future internet testbed. It is also federated with GMOC, a management system in US developed by Indiana Univ., who is a leading research group in this area. Due to federation with GMOC, dvNOC monitors the testbed in the entire world as well.

토픽 모델링을 이용한 인터넷 다형성 웜 탐지규칙 자동 생성 기법

이수철(Suchul Lee),이성일(Sung-il Lee),박경호(Kyungho Park),김성호(Sungho Kim) 한국정보과학회 2014 한국정보과학회 학술발표논문집 Vol.2014 No.12

LDA를 활용한 네트워크 위협 시그니처 추출기법

이성일 ( Sungil Lee ),이수철 ( Suchul Lee ),이준락 ( Jun-rak Lee ),염흥열 ( Heung-youl Youm ) 한국인터넷정보학회 2018 인터넷정보학회논문지 Vol.19 No.1

Network threats such as Internet worms and computer viruses have been significantly increasing. In particular, APTs(Advanced Persistent Threats) and ransomwares become clever and complex. IDSes(Intrusion Detection Systems) have performed a key role as information security solutions during last few decades. To use an IDS effectively, IDS rules must be written properly. An IDS rule includes a key signature and is incorporated into an IDS. If so, the network threat containing the signature can be detected by the IDS while it is passing through the IDS. However, it is challenging to find a key signature for a specific network threat. We first need to analyze a network threat rigorously, and write a proper IDS rule based on the analysis result. If we use a signature that is common to benign and/or normal network traffic, we will observe a lot of false alarms. In this paper, we propose a scheme that analyzes a network threat and extracts key signatures corresponding to the threat. Specifically, our proposed scheme quantifies the degree of correspondence between a network threat and a signature using the LDA(Latent Dirichlet Allocation) algorithm. Obviously, a signature that has significant correspondence to the network threat can be utilized as an IDS rule for detection of the threat.