Physics-Informed Detection of Product-Oriented Attacks in Smart Manufacturing Systems

[works related to this project received an NSF Student Travel Award to attend the 50th NAMRI/SME North American Manufacturing Research Conference (NAMRC 50) at Purdue University, West Lafayette, Indiana]


Production processes are controlled by the laws of physics, and therefore, malicious alterations of products and/or processes intended by cyber-attacks are manifested as anomalous changes in process dynamics. Hence, monitoring physical process variables such as vibration and power consumption (known as side channels in cybersecurity literature) can provide a physical-domain defense layer to detect such attacks. Focusing on product-oriented attacks, I proposed a method to connect product and process designs, and in situ monitoring to identify the physical manifestations of these attacks. The proposed approach can verify the geometric integrity of a machined part by observing cutting power signals during machining. First, I utilized the process and product knowledge to segment the power signal into the cutting cycles corresponding to specific geometrical features and extract process-related information accordingly. This work primarily focuses on extracting machining times for individual geometric features in parts. Next, I used the extracted information to construct quality control charts to use in detecting geometric integrity deviations of machined parts. Finally, I demonstrated the proposed method using a case study of cyber-physical attacks on machining processes aiming to tamper with different product's dimensional and geometrical features.