• 산업경영시스템학회지
• /
• 제21권47호
• /
• pp.139-148
• /
• 1998

CALS IDB (Integrated database) is one of core technologies that embodies the principle of a shared data environment for the life cycle related data in CALS environment. In this study, to successfully share the data, we first classified the data types employed in the CALS environment and then discussed the data heterogeneity issued in data integration processes. To effectively solve this heterogeneity, we proposed the federated database systems as a candidate system especially focusing on the major functions and core element technologies.

• 인터넷정보학회논문지
• /
• 제25권2호
• /
• pp.1-9
• /
• 2024

연합학습은 보안 및 프라이버시 측면에서 중앙 집중식 방법보다 안전하도록 설계되었음에도 불구하고 여전히 많은 취약점을 내재한다. 적대적 공격(adversarial attack)을 수행하는 공격자는 신중하게 제작된 입력 데이터, 즉 적대적 예제(adversarial examples)를 클라이언트의 학습 데이터에 주입하여 딥러닝 모델을 의도적으로 조작하여 오분류를 유도한다. 이에 대한 보편적인 방어 전략은 이른바 적대적 학습(adversarial training)으로 적대적 예제들의 특성을 선제적으로 모델에 학습시키는 것이다. 기존의 연구에서는 모든 클라이언트가 적대적 공격 하에 있는 상황을 가정하는데 연합학습의 클라이언트 수가 매우 많음을 고려하면 실제와는 거리가 있다. 본 논문에서는 클라이언트의 일부가 공격 하에 있는 시나리오에서 적대적 학습의 양상을 실험적으로 살핀다. 우리는 실험을 통해 적대적 예제에 대한 분류 정확도가 증가하면 정상 샘플에 대한 분류 정확도의 감소하는 트레이드오프 관계를 가짐을 밝혔다. 이러한 트레이드오프 관계를 효과적으로 활용하기 위해 클라이언트가 자신이 공격받는지 여부에 따라 손실함수를 적응적으로 선택하여 적대적 학습을 수행하는 방법을 제시한다.

• Mycobiology
• /
• 제49권6호
• /
• pp.551-558
• /
• 2021

The Federated States of Micronesia (FSM) is an island country in the western Pacific and is a known biodiversity hotspot. However, a relatively small number of fungi (236 species) have been reported till July 2021. Since fungi play major ecological roles in ecosystems, we investigated the fungal diversity of FSM from various sources over 2016 and 2017 and constructed a local fungal inventory, which also included the previously reported species. Fruiting bodies were collected from various host trees and fungal strains were isolated from marine and terrestrial environments. A total of 99 species, of which 78 were newly reported in the FSM, were identified at the species level using a combination of molecular and morphological approaches. Many fungal species were specific to the environment, host, or source. Upon construction of the fungal inventory, 314 species were confirmed to reside in the FSM. This inventory will serve as an important basis for monitoring fungal diversity and identifying novel biological resources in FSM.

• 한국정보처리학회:학술대회논문집
• /
• 한국정보처리학회 2023년도 춘계학술발표대회
• /
• pp.582-584
• /
• 2023

연합학습(Federated Learning)에서는 여러 참가자가 서로 간의 데이터를 공유하지 않고 협력하여 하나의 모델을 학습할 수 있다. 그 중 수직 연합학습(Vertical Federated Learning)은 참가자 간에 동일한 샘플에 대해 서로 다른 특성(Feature)를 가지고 학습한다. 또한 서로 다른 특성(Feature)에는 입력의 라벨(Label)도 포함하기 때문에 라벨을 소유한 참가자 외에는 라벨 정보 또한 접근할 수 없다. 이처럼 다양한 참가자가 학습에 참여하는 경우 악의적인 참가자에 의해 모델이 포이즈닝 될 여지가 존재함에도 불구하고 수직 연합학습에서는 관련 연구가 부족하다. 포이즈닝 공격 중 백도어 공격은 학습 과정에 관여하여 특정 입력 패턴에 대해서 모델이 공격자가 원하는 타겟 라벨로 예측하도록 오염시키는 공격이다. 수직 연합학습에서는 참가자가 학습과 추론 모든 과정에서 관여하기 때문에 백도어 공격에 취약할 수 있다. 본 논문에서는 수직 연합학습에서의 최신 백도어 공격과 한계점에 대해 분석한다.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제18권6호
• /
• pp.1462-1477
• /
• 2024

With the evolving complexity of connected vehicle features, the volume and diversity of data generated during driving continue to escalate. Enabling data sharing among interconnected vehicles holds promise for improving users' driving experiences and alleviating traffic congestion. Yet, the unintentional disclosure of users' private information through data sharing poses a risk, potentially compromising the interests of vehicle users and, in certain cases, endangering driving safety. Federated learning (FL) is a newly emerged distributed machine learning paradigm, which is expected to play a prominent role for privacy-preserving learning in autonomous vehicles. While FL holds significant potential to enhance the architecture of the Internet of Vehicles (IoV), the dynamic mobility of vehicles poses a considerable challenge to integrating FL with vehicular networks. In this paper, a novel clustered FL framework is proposed which is efficient for reducing communication and protecting data privacy. By assessing the similarity among feature vectors, vehicles are categorized into distinct clusters. An optimal vehicle is elected as the cluster head, which enhances the efficiency of personalized data processing and model training while reducing communication overhead. Simultaneously, the Local Differential Privacy (LDP) mechanism is incorporated during local training to safeguard vehicle privacy. The simulation results obtained from the 20newsgroups dataset and the MNIST dataset validate the effectiveness of the proposed scheme, indicating that the proposed scheme can ensure data privacy effectively while reducing communication overhead.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제18권5호
• /
• pp.1223-1237
• /
• 2024

Autonomous vehicles use onboard sensors to sense the surrounding environment. In complex autonomous driving scenarios, the detection and recognition capabilities are constrained, which may result in serious accidents. An efficient way to enhance the detection and recognition capabilities is establishing collaborations with the neighbor vehicles. However, the collaborations introduce additional challenges in terms of the data heterogeneity, communication cost, and data privacy. In this paper, a novel personalized federated learning framework is proposed for addressing the challenges and enabling efficient collaborations in autonomous driving environment. For obtaining a global model, vehicles perform local training and transmit logits to a central unit instead of the entire model, and thus the communication cost is minimized, and the data privacy is protected. Then, the inference similarity is derived for capturing the characteristics of data heterogeneity. The vehicles are divided into clusters based on the inference similarity and a weighted aggregation is performed within a cluster. Finally, the vehicles download the corresponding aggregated global model and train a personalized model which is personalized for the cluster that has similar data distribution, so that accuracy is not affected by heterogeneous data. Experimental results demonstrate significant advantages of our proposed method in improving the efficiency of collaborative perception and reducing communication cost.

• 한국패류학회지
• /
• 제29권4호
• /
• pp.343-376
• /
• 2013

The Federated States of Micronesia consists of four states of Yap, Chuuk, Pohnpei and Kosrae, which are located in the West Pacific Ocean. In order to investigate molluscan fauna of Kosrae Island, field survey was made twice from 21st to 30th of January, 2011 and from 6th to 17th of January, 2012 for four localities including 10 intertidal and 14 subtidal zones of Kosrae Island. The mollusk samples collected were identified based on their morphological characteristics, comprising a total of 120 species from 30 families through this survey. In this study, we provided species list and illustrations for 120 species identified, and combined these with the previous record, resulting in a total of 208 species from 47 families in Kosrae Island.

• 한국항공운항학회지
• /
• 제24권2호
• /
• pp.1-10
• /
• 2016

A GPS/SDINS/ST(Star Tracker) integrated sensor algorithm is more robust than the GPS/SDINS and the ST/SDINS systems on exploration of other planets. Most of the advanced studies shown that GPS/SDINS/ST integrated sensor with centralized Kalman filter was more accurate than those 2 integrated systems. The system, however, consist of a single filter, it is vulnerable to defects on failed data. To improve the problem, we work out a study using federated Kalman filter(No-Reset mode) and centralized Kalman filter with adaptive measurement fusion which known as robustness on fault. The simulation results show that the debasing influences are reduced and the computation is enable at least 100Hz. Further researches that the initial calibration in accordance with observability and applying the exploration trajectory are needed.

• 제어로봇시스템학회논문지
• /
• 제5권6호
• /
• pp.764-776
• /
• 1999

In this paper, a tracking algorithm for a maneuvering single target in the presence of multiple data from multiple sensors is investigated. Allowing individual sensors to function by themselves, the estimates from individual sensors on the same target are fused for the purpose of improving the state estimate. The filtering method adopted in the local sensors is the variable dimensional filter with input estimatio technique, which consists of a constant velocity model and a constant acceleration model. A posteriori probability for the maneuvering hypothesis is newly derived. It is shown that the relation function of the a posteriori probability is a function of only the covariance of the fused estimates. Simulation results are provided.

• Animal Systematics, Evolution and Diversity
• /
• 제30권3호
• /
• pp.196-205
• /
• 2014

This paper describes six new species of sponges in the genus Agelas from Kosrae Island, The Federated States of Micronesia. Most Agelasid sponges are known from only tropical regions. All the new Agelas species; A. fragum n. sp., A. kosrae n. sp., A. purpurea n. sp., A. bakusi n. sp., A. vansoesti n. sp. and A. incrustans n. sp. are compared with other valid species that were studied. Six new species differ from the other species by morphology, growth form, skeletal fibres, habitats and spicule size. Agelas fragum n. sp. is characterized by its tuberculate surface and primary fibres with brush-like spicules. Agelas kosrae n. sp. is differs in skeletal structure and have tertiary fibres. Agelas purpurea n. sp. is characterized by primary, secondary and tertiary fibres are all cored with spicules. Agelas bakusi n. sp. is similar to Agelas clathrodes in shape, but differs in the primary fibres. Agelas vansoesti n. sp. is characterized by having acanthostrongyles. Agelas incrustans n. sp. is distinguished by its encrusting and not cavernous interior.