Manufacturing complex parts (e.g. vehicle engines, bionic limbs) at scale has been and still remains a complex task, especially when the fidelity of shapes and material directly translates to their function and performance. For example, if a manufacturing company decided to add a factory in another continent, how would it ascertain that the new factory parts were on par with the parts from its domestic factory? One such solution is the use of additive manufacturing (AM) technology, also known as 3D printing.
Given digital descriptions and print filament (the material used to create the print), AM can produce high fidelity three dimensional objects on demand. Furthermore, AM is becoming increasingly convenient, cost effective, and accurate. As a result, AM has found its way to multitudes of industries including entertainment, transportation, and medicine. However, as with many emerging technologies, it is imperative to assess its susceptibility to malicious intent. What would happen for example, if attackers managed to discreetly modify prosthetic limbs to become far less effective? How would one verify the integrity of a print result?
Dr. Saman Zonouz, a professor at Georgia Tech, sought to address this critical security question. Saman and his collaborators knew that a good evaluation scheme would be one that checked the physical printing process and final product itself, as opposed to its digital source. Tampering of prints may be done without modification to these source files; for example, malware may have been embedded in the print machine or the machine itself may have been physically altered to malfunction. Approaching it from a physical standpoint, Dr. Saman and his team used three channels of information: acoustics, spatial visualizations, and materials. The acoustic and visualization channels were used to verify printing patterns against expected ground-truths—deviation in any of the two channels would strongly hint at unexpected modifications of a print. Finally, the material channel involved the embedding of nanoparticles into the print filament in order to uniquely and discreetly tag the material.
With this new verification proposal, Dr. Saman and his team were able to create a robust scheme, reliant on multiple modes of information instead of having a potential single point of failure. Prior to this paper, most AM research was looking at AM prints’ susceptibility to being stolen from adversaries. As a result, this paper went on to inspire more research in AM print integrity schemes.
While the security of AM is one of Saman’s research interests, he also works with power grid security, intrusion detection systems, reverse engineering, and embedded security verification. Much of his work involves the security of cyberphysical systems where he oftentimes finds himself at the intersection of many different domains. In fact, Saman claimed that it was this interdisciplinary aspect of cyberphysical systems that got him interested in its research. As a new professor at Georgia Tech, Saman is interested in tapping into the expertise of many different domain experts and bridging security gaps. For example, biomedical engineers seldom take into account the perspectives of potential adversaries. Saman is highly interested in identifying these pitfalls and developing their much needed countermeasures.