Guangzhou Panyu Polytechnic, Guangzhou 511400, China.
Comput Intell Neurosci. 2022 Jul 1;2022:2636877. doi: 10.1155/2022/2636877. eCollection 2022.
Training simulators have been gradually evolving in the direction of software, virtualization, networking, and multiplatform in recent years, with the continuous development of hardware and software technologies, particularly the maturity of VR (Virtual Reality) related technologies. The network intrusion program allows remote hackers to take control of the system, posing a serious threat to the network and computer security. As a result, this paper proposes a VR-based ID (intrusion detection) simulation training system. This paper proposes an ID model based on CNN (Convolutional Neural Networks) and LSTM to address these issues (Long Short-term Memory Networks). This model oversamples data from unbalanced data sets, reducing the difference in category data and thus improving the ID model's performance and existing detection methods to compensate for the flaw. 3DSMAX technology is used to simulate the process visualization scene, as well as some key equipment models and signal transmission simulations, during the system design and implementation process. The experimental results show that CNN LSTM outperforms BP, and the overall evaluation index 1 has significantly improved, particularly the 1 index of D4. CNN LSTM outperforms GA (genetic algorithm) by 12.75 percent and BP by 14.07 percent. The system essentially accomplishes the anticipated simulation training goal, and the simulation training effect is impressive.
近年来，随着硬件和软件技术的不断发展，特别是虚拟现实（VR）相关技术的成熟，培训模拟器逐渐朝着软件化、虚拟化、网络化和多平台化的方向发展。网络入侵程序允许远程黑客控制系统，对网络和计算机安全构成严重威胁。因此，本文提出了一种基于 VR 的入侵检测（ID）模拟培训系统。本文提出了一种基于 CNN（卷积神经网络）和 LSTM（长短期记忆网络）的 ID 模型来解决这些问题。该模型对不平衡数据集的数据进行了过采样，减少了类别数据的差异，从而提高了 ID 模型的性能和现有的检测方法来弥补缺陷。在系统设计和实现过程中，使用 3DSMAX 技术模拟了过程可视化场景，以及一些关键设备模型和信号传输模拟。实验结果表明，CNN LSTM 优于 BP，整体评价指标 1 有显著提高，特别是 D4 的 1 指标。CNN LSTM 比 GA（遗传算法）高出 12.75%，比 BP 高出 14.07%。该系统本质上实现了预期的模拟培训目标，模拟培训效果令人印象深刻。
