The Industrial Internet of Things (IIoT), consisting of a large number of self-organized sensors, is one of the prominent representatives of zero touch networks, which will be widely used for information interconnection. With the advancement in intelligent manufacturing, the security of zero touch IIoT becomes a critical issue in various applications. One of the main factors that endanger the normal operation of zero touch IIoT is the weak electromagnetic interference (WEMI) attack, making special precautions necessary for zero touch IIoT. In real-life applications, sensors will be injected with a specific type of noise due to the unique manufacturing process and environment. This noise can be considered as the finger-print of the sensor, which is stable under normal conditions unless the sensor experiences a WEMI attack. Hence, a deep-learning-based WEMI intrusion detection method is employed in this study. First, we introduce the application of Kalman and moving average filters in the fingerprint extraction stage. Second, the frequency and time domain features were extracted from the fingerprint. Third, deep learning models are applied to intrusion detection, and a cloud-edge-end computing framework is proposed. Finally, the experiment analyzes the performance of the WEMI intrusion detection method.

中文翻译：

---

基于深度学习的工业物联网零接触网络弱电磁入侵检测方法


由大量自组织传感器组成的工业物联网（IIoT）是零接触网络的突出代表之一，将广泛应用于信息互联。随着智能制造的进步，零接触工业物联网的安全性成为各种应用中的关键问题。危及零接触工业物联网正常运行的主要因素之一是弱电磁干扰（WEMI）攻击，因此零接触工业物联网需要采取特殊的预防措施。在现实应用中，由于独特的制造工艺和环境，传感器会被注入特定类型的噪声。这种噪声可以被视为传感器的指纹，在正常情况下是稳定的，除非传感器遭受 WEMI 攻击。因此，本研究采用基于深度学习的WEMI入侵检测方法。首先，我们介绍卡尔曼和移动平均滤波器在指纹提取阶段的应用。其次，从指纹中提取频域和时域特征。第三，将深度学习模型应用于入侵检测，提出云边端计算框架。最后通过实验分析了WEMI入侵检测方法的性能。