Nowadays, we are in a typical information age, and network and information security systems are facing severe challenges. However, there is currently no suitable method for the detailed analysis and protection of Internet information security. In order to analyze Internet security information more accurately, we proposed an improved KPCA (Kernel Principle Component Analysis) algorithm within the context of the big data era and in view of the shortcomings and disadvantages of the KPCA feature extraction algorithm. The proposed algorithm not only retains its performance ability, but also improves the subsequent classification ability. This paper uses the KDDCUP99 security audit data set to simulate network intrusions, and the data set used network information data resources within a total of 9 weeks. The training data set contains a total of 7 weeks of data information, and the other 2 weeks of data information are used as a validation data set. The training data set contains a total of 5 million records of network security information, while the verification data set contains 2 million records. The experimental results show that for the network intrusion classification test, the improved algorithm is more efficient, convenient, and faster than the traditional KPCA one. Furthermore, the simulation results also show that the proposed algorithm has achieved a very high degree of accuracy and improvement in terms of “accuracy rate,” “false alarm rate,” and “missing alarm rate.”

如今，我们正处于典型的信息时代，网络和信息安全系统面临严峻挑战。但是，目前没有合适的方法来详细分析和保护Internet信息安全。为了更准确地分析Internet安全信息，鉴于大数据时代的特征和缺点，我们提出了一种改进的KPCA（内核主成分分析）算法。所提出的算法不仅保留了其性能，而且提高了后续的分类能力。本文使用KDDCUP99安全审核数据集来模拟网络入侵，并且该数据集在总共9周内使用了网络信息数据资源。训练数据集总共包含7周的数据信息，其他2周的数据信息用作验证数据集。培训数据集总共包含500万条网络安全信息记录，而验证数据集则包含200万条记录。实验结果表明，对于网络入侵分类测试，改进后的算法比传统的KPCA算法更加高效，便捷和快速。此外，仿真结果还表明，所提出的算法在“准确率”，“虚警率”和“遗漏率”方面都取得了很高的准确性和改进。培训数据集总共包含500万条网络安全信息记录，而验证数据集则包含200万条记录。实验结果表明，对于网络入侵分类测试，改进后的算法比传统的KPCA算法更加高效，便捷和快速。此外，仿真结果还表明，所提出的算法在“准确率”，“虚警率”和“遗漏率”方面都取得了很高的准确性和改进。培训数据集总共包含500万条网络安全信息记录，而验证数据集则包含200万条记录。实验结果表明，对于网络入侵分类测试，改进后的算法比传统的KPCA算法更加高效，便捷和快速。此外，仿真结果还表明，所提出的算法在“准确率”，“虚警率”和“遗漏率”方面都取得了很高的准确性和改进。