The validation technique is independent of printer firmware and software in the controlling computer.

The verification and intrusion detection research will be described August 18 at the 26_th USENIX Security Symposium in Vancouver, British Columbia. The two institutions recently received a grant from the National Science Foundation to further develop the process to be described at the symposium.

These 3-D printed components will be going into people, aircraft and critical infrastructure systems. Malicious software installed in the printer or control computer could compromise the production process. We need to make sure that these components are produced to specification and not affected by malicious actors or unscrupulous producers.

—Raheem Beyah, Georgia Tech

The three components of the new system include:

• Acoustic measurement of the 3-D printer in operation. When compared to a reference recording of a correct print, this acoustic monitoring—done with an inexpensive microphone and filtering software—can detect changes in the printer’s sound that may indicate installation of malicious software.

• Physical tracking of printer components. To create the desired object, the printer’s extruder and other components should follow a consistent mechanical path that can be observed with inexpensive sensors. Variations from the expected path could indicate an attack.

• Detection of nanorods in finished components. Using Raman Spectroscopy and computed tomography (CT), the researchers were able to detect the location of gold nanorods that had been mixed with the filament material used in the 3-D printer. Variations from the expected location of those particles could indicate a quality problem with the component. The variations could result from malicious activity, or from efforts to conserve printer materials.

The researchers tested their technique on three different types of 3-D printers and a computer numerical control (CNC) machine using a polyethylene tibial knee prosthesis as a test case. Beyond detecting malicious activity or quality problems, the technique could stop inadvertent production problems, reducing materials waste.

In their technique to detect flaws in 3-D printed components, the researchers were inspired to apply the same kind of contrast agents used in medical imaging techniques for detecting tumors, said Mehdi Javanmard, assistant professor in the Department of Electrical and Computer Engineering at Rutgers University.

The gold contrast materials were tested to make sure they wouldn’t compromise the structural integrity of the printed components.

Now that they’ve demonstrated the feasibility of the techniques, the researchers plan to use the NSF funding awarded August 1 to improve the validation methods and move them closer to application.

Among the challenges ahead will be obtaining good acoustic data in the noisy environments in which 3-D printers typically operate. In the research reported by the researchers, operation of other 3-D printers near the one being observed cut the accuracy significantly, but Beyah believes that challenge can be addressed with additional signal processing. The technique will also be applied to additional types of printers, and to different materials.

Resources

• Christian Bayens, Tuan Le, Luis Garcia, Raheem Beyah, Mehdi Javanmard and Saman Zonouz, “See No Evil, Hear No Evil, Feel No Evil, Print No Evil? Malicious Fill Patterns Detection in Additive Manufacturing,” (26th USENIX Security Symposium, August 18, 2017).