基于对称不确定性网络流量特征筛选方法
A Feature Filtering Method for Network Flow Based on Symmetric Uncertainty

网络流量特征越来越多，导致相关深度学习模型的训练耗时增加，且对这些特征的选择仍未达成共识。为解决过滤排序式方法中特征子集维数不好确定、包裹式方法中计算耗时过大的问题，本文提出将特征筛选效果良好的基于对称不确定性的过滤排序式方法扩展至过滤式方法中，并设计目标函数兼顾特征与标签之间的相关性和特征之间的冗余度。从初始化和搜索策略两个方面改进用于寻优目标函数的灰狼优化算法，以提高网络流量特征筛选精度。实验结果表明，该方法在保证异常流量检测精度的前提下，取得了最好的特征筛选效果。
More and more network flow features lead to an increase in the training time of related deep learning models. There is still no consensus on the selection of these features. In order to solve the problem that the dimension of the feature subset of the filter-based feature ranking method is not easy to determine, and the calculation time of the wrapper-based feature subset selection method is too large, we propose to extend the filter-based feature ranking method based on symmetric uncertainty, which has a good feature filtering effect, to the filter-based feature subset selection method. A new objective function is designed to take into account the correlation between features and labels and the redundancy between features. The grey wolf optimizer for optimizing the objective function is improved from two aspects of initialization and search strategy to improve the accuracy of network flow feature filtering. The experimental results show that the method proposed in this paper achieves the best feature filtering effect on the premise of ensuring the detection accuracy of abnormal traffic.