The proliferation of industrial cyber physical systems (CPSs) is changing our lives. CPS applications are often associated with sensitive data, core infrastructures, and assets, making them attractive in terms of vulnerability, data breach, and denial of services. Moreover, the heterogeneity in terms of protocols, operating systems, and devices combined with poor adoption of standard solutions create insecure design, architectures, and deployments. In addition, due to the use of wireless technologies, secure communication is strongly needed to protect valuable information. Therefore, secure communication management has become a crucial aspect of developing trustworthy systems with the preservation of security and privacy for CPSs. Deep learning (DL) has strong potential to overcome this challenge via data-driven solutions and improve the performance of CPSs while utilizing limited spectrum resources. DL is a more powerful method of data exploration to learn about “normal” and “abnormal” behavior according to how CPSs' components and devices interact with one another. The input data of each part of a CPS can be collected and investigated to determine normal patterns of interaction, thereby identifying malicious behavior at early stages. Moreover, DL can be important in predicting new attacks, which are often mutations of previous attacks, because they can intelligently predict future unknown attacks by learning from existing examples. Consequently, CPSs must have a transition from merely facilitating secure communication among devices to security-based intelligence enabled by DL methods for effective and secure systems.

中文翻译：

---

客座社论：深度学习驱动的网络物理系统安全通信


工业网络物理系统 (CPS) 的激增正在改变我们的生活。 CPS 应用程序通常与敏感数据、核心基础设施和资产相关联，这使得它们在漏洞、数据泄露和拒绝服务方面具有吸引力。此外，协议、操作系统和设备方面的异构性，加上标准解决方案的采用不佳，导致设计、架构和部署不安全。此外，由于无线技术的使用，强烈需要安全通信来保护有价值的信息。因此，安全通信管理已成为开发可信系统并保护 CPS 安全和隐私的重要方面。深度学习 (DL) 具有强大的潜力，可以通过数据驱动的解决方案克服这一挑战，并在利用有限的频谱资源的同时提高 CPS 的性能。深度学习是一种更强大的数据探索方法，可以根据 CPS 组件和设备之间的交互方式来了解“正常”和“异常”行为。可以收集和研究 CPS 各部分的输入数据，以确定正常的交互模式，从而在早期阶段识别恶意行为。此外，深度学习对于预测新的攻击（通常是先前攻击的突变）非常重要，因为它们可以通过从现有示例中学习来智能地预测未来的未知攻击。因此，CPS 必须从仅仅促进设备之间的安全通信过渡到通过深度学习方法实现基于安全的智能，以实现有效且安全的系统。