Earticle

To overcome the poor communication environment between the underwater acoustic signal transmitter and receiver, previous research applied underwater acoustic OFDM communication system which had efficient bandwidth capacity and could mitigate multi-path delay spread effect via guard interval. Concurrently with this OFDM signal, additional data could be transmitted through spread spreading technique, which was called water-line technique. However, this decreased the performance of the OFDM modulation because the water-line signal degraded OFDM signal. In this paper, we studied the efficient method to minimize signal interference by water-line signal so that we applied convolution error correction code on water-line to minimize signal level.

악성코드는 탐지 프로그램을 피해 기기들에게 피해를 유발한다. 기존의 악성코드 탐지 기법으로 이러한 새로운 악성코드를 탐지하는데 어려움을 겪는 이유는 서명 기반의 탐지 기법을 사용하기 때문이다. 이 기법은 기존 악성코드들 은 효과적으로 탐지하지만, 새로운 악성코드에 대해서는 탐지가 어렵다. 이러한 문제점을 인식하여, 휴리스틱 기법을 추 가적으로 사용한다. 이 논문에서는 딥러닝을 활용하여 악성코드를 탐지하는 기술에 대해 소개하여 새로운 악성코드를 탐지하는 기술에 대해서 제안한다. 또한, 악성코드를 탐지한다는 것은, 기기에서 실행 가능한 파일의 개수는 무수히 많으 므로, 지도학습 방식(Supervisor Learning)으로는 분명한 한계가 존재한다. 그렇기 때문에, 준지도 학습으로 알려진 SCAE(Stacked Convolution AutoEncoder)를 활용한다, 파일들의 바이트 정보들을 추출하여, 이미지화를 진행하고, 이 이미지들을 학습을 시켜, 학습 시키지 않은 10,869개의 악성코드, 3,442개의 비악성코드를 모델에 추론한 결과 정확 도를 98.84%을 달성하였다.

Malicious codes cause damage to equipments while avoiding detection programs(vaccines). The reason why it is difficult to detect such these new malwares using the existing vaccines is that they use “signature-based” detection techniques. these techniques effectively detect already known malicious codes, however, they have problems about detecting new malicious codes. Therefore, most of vaccines have recognized these drawbacks and additionally make use of “heuristic” techniques. This paper proposes a technology to detecting unknown malicious code using deep learning. In addition, detecting malware skill using Supervisor Learning approach has a clear limitation. This is because, there are countless files that can be run on the devices. Thus, this paper utilizes Stacked Convolution AutoEncoder(SCAE) known as Semi-Supervisor Learning. To be specific, byte information of file was extracted, imaging was carried out, and these images were learned to model. Finally, Accuracy of 98.84% was achieved as a result of inferring unlearned malicious and non-malicious codes to the model.

A new method of dose calculation algorithm, called GPU-accelerated Monte Carlo and collapsed cone Convolution (GMCC) was developed to improve the calculation speed of BNCT treatment planning system. The GPU-accelerated Monte Carlo routine in GMCC is used to simulate the neutron transport over whole energy range and the Collapsed Cone Convolution method is to calculate the gamma dose. Other dose components due to alpha particles and protons, are calculated using the calculated neutron flux and reaction data. The mathematical principle and the algorithm architecture are introduced. The accuracy and performance of the GMCC were verified by comparing with the FLUKA results. A water phantom and a head CT voxel model were simulated. The neutron flux and the absorbed dose obtained by the GMCC were consistent well with the FLUKA results. In the case of head CT voxel model, the mean absolute percentage error for the neutron flux and the absorbed dose were 3.98% and 3.91%, respectively. The calculation speed of the absorbed dose by the GMCC was 56 times faster than the FLUKA code. It was verified that the GMCC could be a good candidate tool instead of the Monte Carlo method in the BNCT dose calculations.