Privacy-Enhancing Technologies in Knowledge Engineering for Sensitive Data Applications in AI-Enabled Intelligent Engineering Systems

Abstract

The proliferation of artificial intelligence in intelligent engineering systems has created a highly connected network physical system, which can optimize itself and make real-time decisions. The knowledge engineering pipeline for constructing digital twins, fault diagnosis ontology and adaptive process model usually deals with sensitive data, company data and personal data. This review discusses privacy protection technologies: differential privacy, federated learning, homomorphic encryption and secure multi-party computing. They are an important part of knowledge engineering and can protect the privacy of intelligent engineering systems driven by artificial intelligence. This analysis uses the life cycle principle in system engineering, the feedback theory of network physical system, and the information theory formula about the trade-off between privacy and practicality. Specific integration is evaluated, including digital twins for continuous synchronization, reinforcement learning for dynamic planning, computer vision for online fault detection, and edge computing for deterministic control, and the deployment in real life is considered. The application of intelligent manufacturing, medical care network physical system and intelligent transportation shows great progress in collective intelligence and work efficiency, but there are still problems such as precision loss, long calculation time and security loopholes. Our new idea is to use a control method to treat privacy noise as a minor problem, and to create a system that uses reinforcement learning to select the best method to protect data. We believe that to use these technologies confidently, we need to protect the standard rules and models of privacy from the beginning, and improve the current standards of digital twins and systems engineering.