• 정보처리학회논문지:소프트웨어 및 데이터공학
• /
• 제12권2호
• /
• pp.99-110
• /
• 2023

본 연구는 현 보안 관제 시스템이 직면한 실시간 트래픽 탐지 문제를 해결하기 위해 사이버 위협 프레임워크인 마이터 어택과 머신러닝을 이용하여 유해 네트워크 트래픽을 분류하는 방안을 제안하였다. 마이터 어택 프레임워크에 네트워크 트래픽 데이터셋인 UNSW-NB15를 적용하여 라벨을 변환 후 희소 클래스 처리를 통해 최종 데이터셋을 생성하였다. 생성된 최종 데이터셋을 사용하여 부스팅 기반의 앙상블 모델을 학습시킨 후 이러한 앙상블 모델들이 다양한 성능 측정 지표로 어떻게 네트워크 트래픽을 분류하는지 평가하였다. 그 결과 F-1 스코어를 기준으로 평가하였을 때 희소 클래스 미처리한 XGBoost가 멀티 클래스 트래픽 환경에서 가장 우수함을 보였다. 학습하기 어려운 소수의 공격클래스까지 포함하여 마이터 어택라벨 변환 및 오버샘플링처리를 통한 머신러닝은 기존 연구 대비 차별점을 가지고 있으나, 기존 데이터셋과 마이터 어택 라벨 간의 변환 시 완벽하게 일치할 수 없는 점과 지나친 희소 클래스 존재로 인한 한계가 있음을 인지하였다. 그럼에도 불구하고 B-SMOTE를 적용한 Catboost는 0.9526의 분류 정확도를 달성하였고 이는 정상/비정상 네트워크 트래픽을 자동으로 탐지할 수 있을 것으로 보인다.

• 응용통계연구
• /
• 제22권3호
• /
• pp.541-551
• /
• 2009

Generalized processor sharing(GPS) service policy is a scheduling algorithm to allocate the bandwidth of a queueing system with multi-class input traffic. In a queueing system with single-class traffic, the stationary queue length becomes larger stochastically when the bandwidth (i.e. the service rate) of the system decreases. For a given GPS server, we consider the similar problem to this. We define the monotonicity for the head of the line processor sharing(HLPS) servers in which the units in the heads of the queues are served simultaneously and the bandwidth allocated to each queue are determined by the numbers of units in the queues. GPS is a type of monotonic HLPS. We obtain the HLPS server whose queue length of a class stochastically bounds upper that of corresponding class in the given monotonic HLPS server for all classes. The queue lengths process of all classes in the obtained HLPS server has the stationary distribution of product form. When the given monotonic HLPS server is GPS server, we obtain the explicit form of the stationary queue lengths distribution of the bounding HLPS server. Numerical result shows how tight the stochastic bound is.

• 전기학회논문지
• /
• 제59권4호
• /
• pp.820-822
• /
• 2010

Mobile WiMAX system provides various classes of traffic such as real-time and non-realtime services. These services have different QoS requirements and the QoS aware scheduling has been an important issue. Although many of scheduling algorithms for various services in OFDMA system have been proposed, it is needed to be modified to be applied to Mobile WiMAX system. Since Mobile WiMAX supports five kinds of service classes, it is important to take QoS characteristics of each class into consideration. In this paper, we propose an efficient packet scheduling algorithm to support QoS of each class. Proposed scheme selects a service class first considering QoS Characteristics of each class and choose an appropriate user in the selected class. Simulation results show that the proposed algorithm has better performance than the other algorithm.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제3권2호
• /
• pp.134-146
• /
• 2009

Traffic classification seeks to assign packet flows to an appropriate quality of service(QoS) class based on flow statistics without the need to examine packet payloads. Classification proceeds in two steps. Classification rules are first built by analyzing traffic traces, and then the classification rules are evaluated using test data. In this paper, we use self-organizing map and K-means clustering as unsupervised machine learning methods to identify the inherent classes in traffic traces. Three clusters were discovered, corresponding to transactional, bulk data transfer, and interactive applications. The K-nearest neighbor classifier was found to be highly accurate for the traffic data and significantly better compared to a minimum mean distance classifier.

• 한국경영과학회지
• /
• 제12권1호
• /
• pp.10-18
• /
• 1987

This paper discusses algorithmic properties of a class of complementarity programs involving strictly diagonally isotone and off-diagonally isotone functions, i. e., functions whose Jacobian matrices have positive diagonal elements and nonnegative off-diagonal elements, A typical traffic equilibrium under elastic demands is cast into this class. Algorithmic properties of these complementarity problems, when a Jacobi-type iteration is applied, are investigated. It is shown that with a properly chosen starting point the generated sequence are decomposed into two converging monotonic subsequences. This and related will be useful in developing solution procedures for this class of complementarity problems.

• Geomechanics and Engineering
• /
• 제36권1호
• /
• pp.51-69
• /
• 2024

Currently, composite lining mountain tunnels in China are generally classified based on the [BQ] method for the surrounding rock grade. Increasingly, tunnel field construction is replicated indoors for scale down model tests. However, the development of analogous materials for model tests of composite lining tunnels with different surrounding rock grades is still unclear. In this study, typical Class III and V surrounding rock analogous materials and corresponding composite lining support materials were developed. The whole processes of excavation-support dynamics of the mountain tunnels were simulated. Data on the variation of deformations, contact pressures and strains on the surrounding rock were obtained. Finally, a comparative analysis between model tests and numerical simulations was performed to verify the rationality of analogous material development. The following useful conclusions were obtained by analyzing the data from the tests. The main analogous materials of Class III surrounding rock are barite powder, high-strength gypsum and quartz sand with fly ash, quartz sand, anhydrous ethanol and rosin for Class V surrounding rock. Analogous materials for rockbolts, steel arches are replaced by aluminum bar and iron bar respectively with both shotcrete and secondary lining corresponding to gypsum and water. In addition, load release rate of Class V surrounding rock should be less than Class III surrounding rock. The fenestration level had large influence on the load sharing ratio of the secondary lining, with a difference of more than 30%, while the influence of the support time was smaller. The Sharing ratios of secondary lining in Class III surrounding rock do not exceed 12%, while those of Class V surrounding rock exceed 40%. The overall difference between the results of model tests and numerical simulations is small, which verifies the feasibility of similar material development in this study.

• 전자공학회논문지A
• /
• 제32A권12호
• /
• pp.9-24
• /
• 1995

In this paper, we proposed and analysed two kinds of priority control mechanism to archive the cell loss rate requirement and the delay requirement of each class. The service classes of our concern are the high time priority class(class 1) and the high loss priority class(class 2). Two kinds of priority control mechanism is divided by the method of storing the arriving class 2 cell in buffer on case of buffer full. The first one is the method which discarding the arriving class 2 cell, the second one is the mothod which storing the arriving class 2 cell on behalf of pushing out the class 1 cell in buffer. In the proposed priority schemes, one cell of the class 1 is transmitted whenever the maximum K cells of the class 2 is transmitted on case of transmitting the class 1 cell and the class 2 cell sequentially. In this paper, we analysed the cell loss rate and the mean cell delay for each class of the proposed priority scheme by using the Markov chain. The analytical results show that the characteristic of the mean cell delay becomes better for the class 1 cell and that of the cell loss rate becomes better for the class 2 cell by selecting properly the cell transfer ratio according to the condition of input traffic.

• 융합보안논문지
• /
• 제20권4호
• /
• pp.81-90
• /
• 2020

현실 세계에서는 기존에 알려지지 않은 새로운 유형의 변종 공격이 끊임없이 등장하고 있지만, 인공신경망과 지도학습을 통해 개발된 공격 트래픽 분류모델은 학습을 실시하지 않은 새로운 유형의 공격을 제대로 탐지하지 못한다. 기존 연구들 대부분은 이러한 문제점을 간과하고 인공신경망의 구조 개선에만 집중한 결과, 다수의 새로운 공격을 정상 트래픽으로 분류하는 현상이 빈번하게 발생하여 공격 트래픽 분류성능이 심각하게 저하되었다. 한편, 다중분류 문제에서 각 클래스에 대한 분류가 정답일 확률을 결과값으로 출력하는 소프트맥스(softmax) 함수도 학습하지 않은 새로운 유형의 공격 트래픽에 대해서는 소프트맥스 점수를 제대로 산출하지 못하여 분류성능의 신뢰도 또는 정확도를 제고하는데 한계를 노출하고 있다. 이에 본 논문에서는 소프트맥스 함수의 이러한 특성을 활용하여 모델이 일정 수준 이하의 확률로 판단한 트래픽을 공격으로 분류함으로써 새로운 유형의 공격에 대한 탐지성능을 향상시키는 방안을 제안하고, 실험을 통해 효율성을 입증한다.

• 응용통계연구
• /
• 제20권2호
• /
• pp.219-227
• /
• 2007

시계열 모형은 통신망 트래픽의 예측과 분석에 유용하게 쓰여 왔다. 본 논문에서는 통신망 트래픽의 예측을 위하여 다양한 시계열 모형을 소개하고 성능평가를 하고자 한다. 이를 위하여 실제 통신망 트래픽 자료에 선형 및 비선형 시계열모형을 적합 시키고 비선형 시계열모형이 선형 시계열 모형보다 예측의 정확도가 우수함을 보이고자 한다.

• 한국통신학회논문지
• /
• 제30권9A호
• /
• pp.762-771
• /
• 2005

In this paper, an efficient Adaptive quality-of-service (QoS) traffic control scheme with priority scheduling is proposed for the multimedia traffic transmission over wireless access networks. The objective of the proposed adaptive QoS control (AQC) scheme is to realize end-to-end QoS, to be scalable without the excess signaling process, and to adapt dynamically to the network traffic state according to traffic flow characteristics. Here, the reservation scheme can be used over the wireless access network in order to get the per-flow guarantees necessary for implementation of some kinds of multimedia applications. The AQC model is based on both differentiated service model with different lier hop behaviors and priority scheduling one. It consists of several various routers, access points, and bandwidth broker and adopts the IEEE 802.1 le wireless radio technique for wireless access interface. The AQC scheme includes queue management and packet scheduler to transmit class-based packets with different per hop behaviors (PHBs). Simulation results demonstrate effectiveness of the proposed AQC scheme.