Earticle

최근 클라우드, 빅데이터, 인공지능 등 다양한 분야의 서버 플랫폼이 가상화 기술을 사용하고 있지만 지속적 으로 발생하는 구조적인 보안 취약성이 이슈화 되고 있다. 또한 대부분 가상화 보안 기술은 종류가 제한적이고 플랫폼 제공자에 의존적인 것으로 알려져 있다. 본 논문은 가상화 환경의 안전한 데이터 공유를 위한 다중 인스턴스간 상호인 증 기법에 대해 제안한다. 제안하는 기법은 다중 인스턴스 간에 독립적인 상호인증을 고려하여 보안 구조를 설계하고 키 체인 기법을 적용하여 데이터 공유에 보안 프로토콜의 안전성을 강화 시켰다. 성능분석 결과 기존 보안 구조와는 다른 방식의 안전한 가상화 인스턴스 세션을 생성하고, 상호인증 과정의 각 인스턴스에 대한 효율성이 우수 하다는 것 을 확인하였다.

Recent cloud, big data, there is a problem for the architectural security vulnerability to the server platforms of various fields such as artificial intelligence occurs consistently, but using the virtualization technology. In addition, most secure virtualization technology is known to be dependent on the type is limited and the platform provider. This paper presents a method for mutual authentication for secure data between multiple instances of a shared virtualized environment. The proposed method was designing a security architecture in consideration of the mutual authentication between multiple independent instances, and enhance the safety of a security protocol for sharing data by applying a key chain techniques. Performance analysis results and the existing security architecture demonstrated that protect each virtualized instances of the session and the other way, a compliance effectiveness for each instance of the mutual authentication process.

A novel fuzzy learning framework that employs fuzzy inference to solve the problem of Multiple Instance Learning (MIL) is presented. The framework introduces a new class of fuzzy inference systems called Multiple Instance Mamdani Fuzzy Inference Systems (MI-Mamdani). In multiple instance problems, the training data is ambiguously labeled. Instances are grouped into bags, labels of bags are known but not those of individual instances. MIL deals with learning a classifier at the bag level. Over the years, many solutions to this problem have been proposed. However, no MIL formulation employing fuzzy inference exists in the literature. Fuzzy logic is powerful at modeling knowledge uncertainty and measurements imprecision. It is one of the best frameworks to model vagueness. However, in addition to uncertainty and imprecision, there is a third vagueness concept that fuzzy logic does not address quiet well, yet. This vagueness concept is due to the ambiguity that arises when the data have multiple forms of expression, this is the case for multiple instance problems. In this paper, we introduce multiple instance fuzzy logic that enables fuzzy reasoning with bags of instances. Accordingly, a MI-Mamdani that extends the standard Mamdani inference system to compute with multiple instances is introduced. The proposed framework is tested and validated using a synthetic dataset suitable for MIL problems. Additionally, we apply the proposed multiple instance inference to fuse the output of multiple discrimination algorithms for the purpose of landmine detection using Ground Penetrating Radar.

Even though mainly statistical methods have been used in anomaly network intrusion detection, to detect various attack types, machine learning based anomaly detection was introduced. Machine learning based anomaly detection started from research applying traditional learning algorithms of artificial intelligence to intrusion detection. However, detection rates of these methods are not satisfactory. Especially, high false positive and repeated alarms about the same attack are problems. The main reason for this is that one packet is used as a basic learning unit. Most attacks consist of more than one packet. In addition, an attack does not lead to a consecutive packet stream. Therefore, with grouping of related packets, a new approach of group-based learning and detection is needed. This type of approach is similar to that of multiple-instance problems in the artificial intelligence community, which cannot clearly classify one instance, but classification of a group is possible. We suggest group generation algorithm grouping related packets, and a learning algorithm based on a unit of such group. To verify the usefulness of the suggested algorithm, 1998 DARPA data was used and the results show that our approach is quite useful.

본 연구는 MRE 영상으로부터 크론병 환자의 염증 활성도를 자동으로 평가하기 위해, 파운데이션 모델 DINOv2와 어텐션 기반 다중 인스턴스 학습을 결합한 딥러닝 기법을 제안한다. 제안 기법은 개별 장염증의 MRE 영상 패치를 다중 인스턴스로 처리하고, DINOv2 모델을 특징 추출기로 활용하고 Top-K 어텐션 풀링 전략을 적용하여 핵심 병변에 집중하는 과정을 통해 환자별 sMaRIA 점수를 산출한다. 208명의 환자 데이터에 대한 검증 결과, DINOv2-Small과 어텐션 기반 풀링 및 Top-K 인스턴스 선택(K=3)을 결합한 모델이 77.78%의 가장 우수한 분류 정확도를 달성하였으며, 시각화를 통해 모델이 임상적으로 유의미한 병변에 집중함을 확인하였다.

This study proposes a deep learning framework that combines a foundation model, DINOv2, with an attention-based multiple instance learning (MIL) strategy to automatically assess inflammatory activity in Crohn's disease from magnetic resonance enterography (MRE) images. The proposed method treats lesion-level MRE patches as multiple instances and employs DINOv2 as a feature extractor, followed by a Top-K attention pooling strategy to focus on clinically relevant lesions and derive patient-level sMaRIA scores. Experimental validation on a dataset of 208 patients demonstrated that the model integrating DINOv2-Small with attentionbased pooling and Top-K instance selection (K = 3) achieved the highest classification accuracy of 77.78%. Qualitative results further confirmed that the model effectively attends to clinically meaningful inflammatory regions.

본 논문에서는 파티클 필터를 장착하고 WMIL(Weighted Multiple Instance Learning)을 이용한 전방차량 추적 알고리즘을 제안하였다. 제안된 알고리즘에서 영상표현은 Haar-like 특징들을 사용하고 차량인식 결과는 추적하고자 하는 전방차량의 위치를 알아내는데 사용된다. 제안된 방식에서 WMIL과 파티클 필터를 결합하기 위해 기존의 외관모델을 이용한 추적에서 탐색영역에서 영상조각의 추적객체 신뢰도 맵을 계산하는 대신에 파티클 필터의 전파, 관측, 추정, 선택 그리고 분류기 훈련 등의 단계를 매 프래임 마다 순차적으로 수행하여 객체의 새로운 위치를 갱신하였다. 제안된 전방차량 추적방식은 실험을 통해 Ada-boost, MIL(Multiple Instance Learning)이나 WMIL 방법을 이용하는 추적에 비해 파티클 필터로 인해 계산량 증가는 불가피하나 추적의 질적인 정확도는 국도, 고속도로, 터널 및 시내도로 등의 실험 동영상에서 추적대상의 위치오차가 평균 4.5화소 정도로 기존의 추적방법들에 비해 크게 개선되는 것을 확인하였다.

This paper proposes a novel forward vehicle tracking algorithm based on the WMIL(Weighted Multiple Instance Learning) equipped with a particle filter. In the proposed algorithm Haar-like features are used to train a vehicle object detector to be tracked and the location of the object are obtained from the recognition result. In order to combine both the WMIL to construct the vehicle detector and the particle filter, the proposed algorithm updates the object location by executing the propagation, observation, estimation, and selection processes involved in particle filter instead of finding the credence map in the search area for every frame. The proposed algorithm inevitably increases the computation time because of the particle filter, but the tracking accuracy was highly improved compared to Ababoost, MIL(Multiple Instance Learning) and MIL-based ones so that the position error was 4.5 pixels in average for the videos of national high-way, express high-way, tunnel and urban paved road scene.