선택한 단어 수는 입니다.
최소 단어 이상 선택하여야 합니다.
최소 단어 이상 선택하여야 합니다.
선택한 단어 수는 30입니다.
최대 10 단어까지만 선택 가능합니다.
최대 10 단어까지만 선택 가능합니다.
제안 방법
- For the data set employed for the present study, the data provided by Cert insider threat Center of Carnegie Mellon’s Software Engineering was used, and was checked through the classification algorithms of SVM, NaiveBayes, Multilayerperceptron, and RandomForest to perform clasification.
- However, new attack patterns of diversified forms are occurring today due to the complexity of an orgnization and the changes of systems. Such changes are being made to a study of prediction from the standpoint of defenders or detection, for the studies are being carried out based on the intelligent methods by way of predicting the data convergence technology for the system and the system changes as well as understnding on insiders. In the present chapter, related studies to revent such insider threats are introduced.
- In the present experiment, classifiation for the insider activitys is performed. The classification was realized into normal user and threat user, the transition matrix was constructed by using CERT:insider threat data set, and a total of 4 classification algorithms of SVM, NaiveBayes, Multilayerperceptron, and RandomForest were excuted. For Tranining Data and Test Data, experiments were implemented by application of 10-fold cross-validation[12][13].
- Using this classification, a discussion was made on the problems including diversified vulnerabilities of the IC system model deriving scenarios concerning insiders’ threat and the characteristics of the events related to insider attacks in the attacker model.
대상 데이터
- For Tranining Data and Test Data, experiments were implemented by application of 10-fold cross-validation[12][13]. The whole users of the data set consisted of 1,000 people in total, of whom 30 people were defined as threats. In the makeup of the data set for the present experiment, 170 users were made as the subject by randomly selecting only about 15% of the whole users.
이론/모형
- For the activity analysis, changes in the operation processes for employees’s task states were analyzed by application of the Markov chain model.
- are included. For the anomaly detection algorithm, abnormalities were detected by using the outlier detection algorithm based on KNN (K-nearest neighbor), which was supervised. According to the experimental results, the detection algorithm showed performance of a reliable level through ROC.
- 5, classification for the inside activitys is performed. To perform the classification and verify the classification, 10-fold cross-validation was applied to the training set, and checking was made through the classification algorithms of SVM, NaiveBayes, Multilayerperceptron, and RandomForest. An overall Flow Chart for propsed method is as shown below in Fig.
※ AI-Helper는 부적절한 답변을 할 수 있습니다.