http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
BERT 기반의 모델을 이용한 무기체계 소프트웨어 정적시험 거짓경보 분류 모델 개발 방법 연구
남효주,이인섭,정남훈,정성윤,조규태,노성규 한국정보과학회 2024 정보과학회논문지 Vol.51 No.7
최근 무기체계에서 소프트웨어의 규모와 복잡도가 커짐에 따라 소프트웨어의 신뢰성 및 안정성 확보가 요구되고 있다. 이를 위해 개발자는 정적 및 동적 신뢰성 시험을 수행해야한다. 하지만 정적시험 과정에서 많은 거짓경보들이 발생하여 이를 분석하고 처리하는데 많은 시간과 자원을 할애하고 있다. 기존 연구에서는 이러한 문제를 해결하기 위해 SVM, LSTM 등의 모델을 활용하여 거짓 경보를 분류한다. 하지만 연구들에서 사용된 모델의 입력값은 코드 관련 정보이거나, Word2Vec기반 코드 임베딩이므로 결함 발생 부분과 연관된 코드 간의 관계를 표현하지 못한다는 한계점이 존재한다. BERT기반의 모델은 양방향 트랜스포머의 적용을 통해 문장 간 앞뒤 관계를 학습하므로 코드 간 관계를 분석하는데 용이하다. 따라서 이를 거짓 경보 분류 문제에 활용하면 위 한계점을 극복할 수 있다. 본 논문에서는 정적시험 결과를 효율적으로 분석하기 위해 BERT기반의 모델을 활용한 거짓경보 분류 모델 개발 방법을 제안한다. 개발 환경에서 데이터셋을 구축하는 방법을 설명하고, 실험을 통해 분류 모델의 성능이 우수함을 보인다. Recently, as the size and complexity of software in weapon systems have increased, securing the reliability and stability is required. To achieve this, developers perform static and dynamic reliability testing during development. However, a lot of false alarms occur in static testing progress that cause wasting resources such as time and cost for reconsider them. Recent studies have tried to solve this problem by using models such as SVM and LSTM. However, they have a critical limitation in that these models do not reflect correlation between defect code line and other lines since they use Word2Vec-based code embedding or only code information. The BERT-based model learns the front-to-back relationship between sentences through the application of a bidirectional transformer. Therefore, it can be used to classify false alarms by analyzing the relationship between code. In this paper, we proposed a method for developing a false alarm classification model using a BERT-based model to efficiently analyze static test results. We demonstrated the ability of the proposed method to generate a dataset in a development environment and showed the superiority of our model.
AutoML을 이용한 무기체계 소프트웨어 정적시험 거짓경보 분류 모델 개발
남효주,이인섭,김태우,정남훈,조규태,노성규 한국정보과학회 2023 정보과학회 컴퓨팅의 실제 논문지 Vol.29 No.2
As weapon systems become more sophisticated, most functions are being developed by software (SW) code. With the proportion of SW in weapon systems increasing gradually, its complexity has also increased significantly. Therefore, SW stability has become an important issue in weapon system development. To meet the needs, the Defense Acquisition Program Administration makes SW developers perform SW reliability testing. However, there is a problem in that a large number of false alarms generate on scalable software in the static test of SW reliability testing. So the SW developers have to determine manually whether they are false alarms or not, which leads to overspending. In this paper, we propose a machine learning-based false alarm classification model using AutoML. It shows good performance in terms of four evaluation metrics such as accuracy, recall, precision, and F1-score.
남효주,정선미,김재송,김수현,손은선 한국병원약사회 2019 병원약사회지 Vol.36 No.1
Background : Invasive fungal infections are major infections that increase morbidity and mortality in intensive care units (ICUs). Most infected patients are immunosuppressed or critically ill. In 2016, the Infectious Diseases Society of America (IDSA) guidelines recommended a combination of voriconazole and echinocandins for patients with invasive aspergillosis. The purpose of this study was to analyze the clinical efficacy of combination antifungal therapy in ICUs of tertiary-care hospitals and to provide reference information for future treatment of fungi. Methods : We respectively reviewed the electronic medical records (EMRs) of patients who had been treated with combination antifungal therapy more than three days in ICUs from January 1 to December 31, 2016. To analyze the current status of combination antifungal therapy, we analyzed the type of concomitant medications, the dosage, and the periods of administration. The reasons the drugs were given were categorized into four groups, and efficacy was also assessed. Results : A total of 21 patients were enrolled in this study. The combination of voriconazole and caspofungin was the most administered medication. The average duration of administration was 13.67±16.22 days. One patient (5%) received antifungal treatment for prophylactic purposes, seven patients (33%) were treated for empirical reasons, four patients (19%) for predictive purposes, and nine patients (43%) for therapeutic purposes. The drugs were effective in nine patients (43%), ineffective in four patients (19%), and eight patients (38%) died before assessment. Conclusions : This study confirmed that a combination of antifungal agents was effective in ICU patients. Future research is necessary to further establish effectiveness.
자연어처리 모델을 이용한 무기체계 소프트웨어 정적시험 거짓경보 저감 연구
이인섭,남효주,정남훈,조규태,노성규 한국정보과학회 2024 정보과학회논문지 Vol.51 No.3
Recently, Securing software stability has become increasingly important as military systems have been upgraded. To this end, the Defense Acquisition Program Administration conducts reliability tests for weapon system software through static analysis tools. However, many false alarms occurred during the test process, resulting in a waste of time and resources. This paper aims to achieve a high positive/false positive classification rate by creating a dataset using the log of a static analysis tool and training a language model. Additionally, data processing methods appropriate for the static analysis features of weapon system software were investigated and analyzed during the research. As a result of the analysis, it was found that the CodeBert model pretrained in C/CPP and natural language using Optuna, a hyperparameter tuning tool, showed 4-5% higher performance based on the F1 score than the existing SoTA model. If the model presented in this research is mainly employed in software static testing, a significant number of false positives can be found.
입력 변화에 강건한 소스 코드 취약점 탐지를 위한 LSTM-Attention 모델
정관영,정병길,천세범,남효주,이인섭,정남훈,조규태,이상근 한국정보과학회 2024 정보과학회 컴퓨팅의 실제 논문지 Vol.30 No.3
Source code vulnerability detection is a very important step, and deep learning-based methods are increasingly being used for this purpose. In this study, we train and compare various deep neural network architectures to detect vulnerabilities in the source code. In this work, experimentally, we demonstrate that the LSTM-Attention model is highly effective in vulnerability detection compared to other techniques. We provide significant insights into the consistent performance of this model under varying conditions. In particular, we delve deep into the impact of input data length, embedding techniques, and the size of latent vectors on the performance of the model. This will be an important reference for determining whether the LSTM-Attention model can ensure consistent performance in a variety of real-world situations, and under what conditions it exhibits the best performance.
국제보건에서 한의약 공적개발원조의 현재와 지속가능한 발전전략
황예은,이승현,김형우,남효주,이승언,백유상,채한 대한한의학회 2024 대한한의학회지 Vol.45 No.1
Objectives: Korea has a unique history of being both a recipient and a donor of Official Development Assistance (ODA), and the international community expects Korea to contribute to the development of developing countries by utilizing this experience. Traditional Korean medicine (TKM) seeks to contribute to global health, however the concept of ODA has been unclear and there has been no clear strategy and sustainable initiatives. Methods: This study examines the concept of ODA and its application in global health, including business objectives, scale, evaluation principles, and development strategies. Additionally, we reviewed the current status of Traditional Korean medicine globalization projects and conducted a SWOT analysis of the internal and external environment of the TKM sector. Based on these findings, we redefined the concept of ODA for TKM and proposed suggestions for its development. Results: The current study identified key ideas for TKM ODA. It should prioritize the improvement of primary healthcare in recipient countries, aligning with the international evaluation criteria of the SDGs. Secondly, TKM's 70 years of experience can be leveraged to enhance both the competence and economic benefits of recipient countries' medical systems. Based on these concepts, a TKM ODA development model was proposed, comprising two core visions, three development strategies and goals, and six core values. Conclusion: This study systematically examined the TKM in global health and suggested sustainable development strategies for TKM ODA. Through its expansion, TKM could contribute to the advancement of global traditional medicine and its overall engagement in healthcare.