Attacks on Confidentiality of Additive Manufacturing Systems using Acoustic Side-Channel

Acoustics emanated from the additive manufacturing systems, such as 3D-printers, carry process information that can be used to breach the confidentiality of the system. Moreover, this is an example of a physical-to-cyber domain attack, where information gathered from the physical domain, such as acoustic side-channel of a 3D-printer, can be used to reveal information about the cyber domain (such as the G-code used to operate the 3D-printer). To demonstrate the vulnerability of the system to such type of attacks, we are presenting a novel attack model consisting of signal processing, machine learning algorithms, and context-based post-processing. In our experiments, we have successfully conducted the attack by reconstructing the test object and its corresponding G-code with an accuracy of 89.72%. These attacks on confidentiality can pose serious threat to additive manufacturing systems as it can lead to theft of Intellectual Property (IP) and trade secrets.