Inference Attack and Privacy Security of Data-driven Industrial Process Monitoring Systems

In modern industry, data-driven process monitoring systems (PMS), as the initial defense line of industrial control system security, have been widely used in all walks of life. However, the privacy security of the data-driven PMS itself has rarely or never received serious attention. Once the data-driven PMS suffers from intrusion and malicious attacks, it will not only interfere with the normal operation of the industrial control system, but also lead to the disclosure of industrial confidential and privacy information and major economic losses. To handle this issue, this work proposes a novel pioneering study on the inference attack and privacy security problem in the data-driven PMS. Firstly, the potential attack and privacy violation risks of data-driven PMS are investigated. Second, a novel industrial inference attack and privacy security benchmark on data-driven PMS is presented, in which a series of membership inference attack and defense experiments are designed and conducted. Third, we provided a detailed discussion about which member reasoning attacks are the most potential threats to the data-driven PMS and which defense technologies are most suitable for mitigating the attack. The experimental results will provide researchers and practitioners with a new perspective when designing a novel data-driven PMS with more robust and privacy protection performance.