基于深度序列加权核极限学习的入侵检测算法

针对海量多源异构且数据分布不平衡的网络入侵检测问题以及传统深度学习算法无法根据实时入侵情况在线更新其输出权重的问题，提出了一种基于深度序列加权核极限学习的入侵检测算法（DBN-WOS-KELM算法）。该算法先使用深度信念网络DBN对历史数据进行学习，完成对原始数据的特征提取和数据降维，再利用加权序列核极限学习机进行监督学习完成入侵识别，结合了深度信念网络提取抽象特征的能力以及核极限学习机的快速学习能力。最后在部分KDD99数据集上进行了仿真实验，实验结果表明DBN-WOS-KELM算法提高了对小样本攻击的识别率，并且能够根据实际情况在线更新输出权重，训练效率更高。

Intrusion detection algorithm based on depth sequence weighted kernel extreme learning

This paper proposed a intrusion detection algorithm based on deep sequence weighting kernel limit learning(DBN-WOS-KELM) to solve the problem of massive multi-source heterogeneous network intrusion detection with unbalanced data distribution and the problem, that the traditional deep learning algorithm could not update its output weight online according to the real-time intrusion situation. The algorithm firstly used the deep belief network DBN to study the historical data, then extracted the features of the original data and reduced the dimension of the data. And then used the weighted sequence kernel extreme learning machine for supervised learning to complete the intrusion detection. It combined the ability of extracting abstract features from the deep belief network and the fast learning ability of the kernel extreme learning machine. Finally, the simulation experiments on KDD99 dataset show that DBN-WOS-KELM algorithm improves the recognition rate of small sample attacks, and can update the output weights online according to the real-time situation, so that the training efficiency is much higher.