基于属性相似度云模型的网络异常检测

针对网络异常检测虚警率偏高的问题,提出了一种基于属性相似度云模型的网络异常检测新方法。基于各属性对分类的不同贡献,结合数据对象空间和属性空间的相似度概念,给出了属性相似度和属性权重的计算方法,该方法可降低网络数据空间的维数,提高目标识别的准确率。试验表明,该方法具有先验知识需求少和参数容易确定的优点,能比较准确地检测出对网络数据的异常行为。

Network anomaly detection based on attributes similarity and cloud model

A new method for the anomaly detection based on the attributes similarity and the cloud model was proposed to alleviate the high false positive rate problem in the detection. The method for calculating the attribute similarity and the attribute weight was given based on the different contribution of the attributes to the classification in combination with the concept of the similarity between the data object space and the attribute space. Tile method can reduce the dimensionality of the network data space, improve the veracity of the anomaly detection. The experiment results indicated that the proposed method is characterized by that few a prior knowledge is needed and the parameters are easy to determine. It can enhance the detection performance and reduce its false positive rate.