EthicalFab: Toward ethical fabrication process through privacy-preserving illegal product detection

The rapid evolution of Industry 4.0 and digital manufacturing technologies has enabled data-driven innovation, particularly in additive manufacturing. However, these advancements raise significant ethical concerns regarding product legality and data privacy. For instance, the printing of illegal objects such as guns could compromise public safety. This paper addresses this challenge by detecting and preventing the printing of illegal objects such as guns without leaking private data. Specifically, we propose neural network classifiers with embedded differential privacy (DP) mechanisms for identifying 3D printed objects by analyzing video frames during the fabrication process. We evaluate the classification performance in both offline and sequential settings for binary and multiclass classification tasks. The results demonstrate that our privacy-preserving framework can effectively identify malicious objects while ensuring that privacy is maintained. The proposed system promises to play a crucial role in safeguarding ethical fabrication process. Additionally, the framework paves the way for future work in privacy-preserving machine learning applications in industrial settings. Code is available at https://github.com/minsung-k/EthicalFab.