基于生成式对抗网络的网络安全态势感知

Network Security Situation Awareness Based on Generative Adversarial Networks

摘要:

研究了生成式对抗网络在网络安全态势感知中的应用。采用生成对抗网络的生成器与判别器，分析Sigmoid函数、Tanh函数、ReLU函数、LeakyReLU函数的损失变化，对生成式对抗网络的网络感知损失进行平衡处理。通过离散随机变量的变换，得到输出网络安全数据的生成函数，在生成式对抗网络下，反映网络安全态势瞬间的离散状态，量化大规模离散网络安全态势感知值，从而实现网络安全态势的精确感知。实验验证结果显示，AP2、AP3、AP4在第3个攻击阶段中，态势感知值达到最高点；AP1在第5个攻击阶段中，态势感知值达到最高点，感知结果与实际结果相符，感知性能良好，对于提高网络安全性具有重要作用。

Abstract:

The application of generative adversarial networks in network security situational awareness had been discussed. Using the generator and discriminator of generative adversarial networks, analyze the loss changes of Sigmoid function, Tanh function, ReLU function, and LeakyReLU function, and balance the network perception loss of generative adversarial networks. By transforming discrete random variables, a generation function for outputting network security data is obtained. In a generative adversarial network, the instantaneous discrete state of the network security situation is reflected, and the large-scale discrete network security situation perception value is quantified to achieve accurate perception of the network security situation. The results showed that AP2, AP3, and AP4 reached their highest situational awareness values during the third attack phase; In the fifth attack phase, AP1"s situational awareness value reached its highest point, and the perception results were consistent with the actual results. The perception performance was good and played an important role in improving network security.