作者: Hui-Hao Chou , Sheng-De Wang
DOI: 10.1109/CCST.2015.7389649
关键词:
摘要: As Internet attacks grow rapidly, firewalls or network intrusion systems are indispensable. Existing approaches usually use attack signatures, machine learning data mining algorithms to detect and stop anomaly malicious flow. Machine need a set of labeled train the detection model, while is not always available. In this paper, we proposed an approach that adaptive ever-changing environment. The constructs decision tree-based model for from unlabeled by using unsupervised algorithm called spectral clustering. And system can easily be deployed on cloud experiments with DARPA 2000 KDD Cup 1999 set, our shows notable improvement performance after adaptation procedure.