DT4I4-Secure: Digital Twin Framework for Industry 4.0 Systems Security
DT4I4-Secure: Digital Twin Framework for Industry 4.0 Systems Security
复制标题
DOI：
10.1109/uemcon59035.2023.10316090
复制
发表时间：
2023-10
期刊：
影响因子：
--
通讯作者：
Yu-Zheng Lin;Sicong Shao;Md Habibor Rahman;M. Shafae;Pratik Satam
中科院分区：
文献类型：
--
作者：
Yu-Zheng Lin;Sicong Shao;Md Habibor Rahman;M. Shafae;Pratik Satam
网络物理系统（CPS）自动化的快速采用正在触发工业4.0（I4.0），将云计算，机器学习（ML），人工智能（AI）和通用网络连接集成到传统的孤立系统中。这些I4.0变化正在优化智能制造（SM）系统的性能，但代价是复杂性增加，使I4.0系统面临比以往更多的网络攻击。为了应对这些挑战，这项工作提出了DT 4 I4-Secure：工业4.0安全的数字孪生框架。DT 4 I4-Secure提供了一个框架，使用模型组合（包括物理模型和基于数据的模型）为I4.0系统创建数字孪生（DT）。本文展示了使用DT框架来检测对I4.0系统的攻击，通过比较观察结果与DT的未来预测。本文评估了DT 4 I4-Secure的计算机数控（CNC）车削工艺制造金属阀芯的性能，其中实验结果表明，该模型可以预测正常操作的平均绝对误差（MAE）为0.005081。本工作还探讨了使用指数加权移动平均（EWMA）为基础的动态阈值，而不是传统的静态阈值的攻击检测时，数控车削过程中，在三个单独的攻击场景。结合动态阈值的DT 4 I4-Secure在所有三种攻击场景下的F1分数提高了3.46倍，同时在制造周期内具有100%的准确性。
The rapid adoption of automation in the Cyber-Physical Systems (CPS) is triggering Industry 4.0 (I4.0), integrating cloud computing, machine learning (ML), artificial intelligence (AI), and universal network connectivity into traditionally isolated systems. These I4.0 changes are optimizing the performance of Smart Manufacturing (SM) systems at the cost of increased complexity, exposing I4.0 systems to more cyberattacks than ever before. To address these challenges, this work presents DT4I4-Secure: A Digital Twin Framework for Industry 4.0 Security. The DT4I4-Secure presents a framework to create Digital Twins (DT) for I4.0 systems using a combination of models (including physics and data-based models). This paper showcases the use of the DT framework to detect attacks on I4.0 systems by comparing observations with future predictions from the DT. This paper evaluates the performance of the DT4I4-Secure for a Computer Numerical Control (CNC) turning process manufacturing a metallic spool, wherein the experimental results show the model can predict normal operations with a mean absolute error (MAE) of 0.005081. This work also explores using an Exponentially Weighted Moving Average (EWMA) based dynamic threshold instead of a traditional static threshold for attack detection when the CNC turning process is under three separate attack scenarios. The DT4I4-Secure combined with the dynamic threshold shows a 3.46 times improvement in F1-Scores over all three attack scenarios for instantaneous attack detection while having 100% accuracy during the manufacturing cycle.