Cyber-physical vulnerabilities in additive manufacturing systems: A case study attack on the .STL file with human subjects

Highlights

• An overview of the AM process chain and attack vectors at each step is given.

• The .STL file as an attack vector is investigated in detail.

• An automated attack algorithm is used to insert voids into arbitrary geometries.

• A study using human subjects is conducted to simulate an attack occurring.

Abstract

One of the key advantages of additive manufacturing (AM) is its digital thread, which allows for rapid communication, iteration, and sharing of a design model and its corresponding physical representation. While this enables a more efficient design process, it also presents opportunities for cyber-attacks to impact the physical word. In this paper the authors examine potential attack vectors along the Additive Manufacturing process chain. Specifically, the effects of cyber-physical attacks, and potential means for detecting them, are explored. To explore the potential implications of such an attack, a case study was conducted to evaluate the ability of human subjects to detect and diagnose a cyber-physical attack on the STL file of a test specimen. Based on the results of this study, recommendations are presented for preventing and detecting cyber-physical attacks on AM processes.

Introduction

Cyber-physical systems (CPS) are systems that integrate physical hardware with software systems, often with the use of a network. With the growth of the Internet of Things (IoT) the number of CPS systems on networks continues to increase [1]. Concurrently, cyber-attacks have become more prevalent in recent years, increasing in maliciousness and decreasing in visibility [1], [2], [3]. This poses a significant issue, as cyber-attacks on cyber-physical systems could result in damage to the machines themselves or the humans who interact with them.

A prominent example of a cyber-physical attack was the Stuxnet worm that targeted Iranian centrifuges used for refining uranium. In this attack the worm was able to infect the software system and affect the physical hardware, causing damage to the centrifuges. By sending false data back to the operators, Stuxnet was able to make it appear as though the centrifuges were operating correctly, while it caused them to damage themselves. The ability of Stuxnet both to cause damage to physical systems and to hide itself illustrates the ability of a cyber-physical attack to disrupt manufacturing systems and the need for physical methods of detection [4].

Another example of a cyber-physical attack is the hijacking of insulin pumps. In this case a hacker is able to connect to a Bluetooth enabled insulin pump to control the dose of insulin given to the wearer. By increasing or decreasing the dosage of insulin, it is possible to cause serious injury or even death in the user. The currently security system for these pumps is insufficient to prevent a cyber-physical attack that could have potentially lethal consequences [5].

The previously mentioned examples demonstrate the ability for cyber-attacks to cross over into the physical world. Attacks on CPS are even more alarming when considering the ever-increasing amount of networked devices that are being connected to machines in the manufacturing world. A cyber-attack on these machines could cause injury to plant workers and damage to the machine itself. This has already been demonstrated at a German foundry wherein a cyberattack on a blast furnace’s control systems prevented a safe shut down and resulted in “massive damage” [6], [7]. Perhaps even more insidiously, an attack could be designed to cause a process to produce faulty parts that might find their way into end-user products[8]. For example, an attack could be designed to affect the production of a jet turbine part such that it would pass inspection but fail during operation and cause significant damage.

With the rise in both the number of CPS connected to networks and in malicious cyber-attacks, there is a clear need for research to understand the vulnerabilities of cyber-physical systems. Recent work on the current status of cyber-physical systems has identified security as a major issue [9]. Additionally, there is a need to identify the skills that are necessary to effectively operate a cyber-physical manufacturing environment [10], which the authors believe includes skills pertaining to the identification and prevention of attacks on manufacturing systems.

As such, the authors have investigated cyber-physical vulnerabilities in manufacturing systems. The authors have demonstrated such attacks on subtractive manufacturing, and have found that they can have a demonstrable effect on the part being produced [11], [12]. In the previous work the authors demonstrated a toolpath attack by manually modifying and replacing machining toolpath (e.g., GCODE). This work was limited to modifying the external geometry of a part in a way that affected the part strength, but was detectable using common inspection techniques.

In this paper, the authors scope their research solely on Additive Manufacturing (AM, commonly referred to as “3D Printing”) systems. The process chain of these networked machines has unique vulnerabilities that warrant a detailed investigation. Specifically, due to (i) their layer-wise fabrication approach and (ii) the processes’ impact on parts’ final material properties due to process-induced transformation of raw material to finished good (as opposed to subtractive machining, where the material properties are defined by the feed stock), there are several cyber-physical vulnerabilities that are unique to AM. For example, voids can be placed inside of a part and the material properties of internal layers can be changed without affecting the exterior layers, which makes detection with traditional part inspection techniques difficult (as discussed in Section 3). Because of the potential damage from a cyber-physical attack, there is a need to investigate AM systems to determine what vulnerabilities exist and how to prevent and mitigate the threat of cyber-attacks. Recently, there has been increasing interest in evaluating these threats to AM systems, with the effects of toolpath modifications and embedded defects being simulated and evaluated through physical testing [13]. The key differences between additive and subtractive manufacturing are the features that make AM unique: (i) its layerwise approach to fabrication, which provides opportunity to place defects internally, and (ii) its transformation of raw material in the process, which allows for altering material properties of the final product by changing the print parameters.

An overview of the cyber-physical attack vulnerabilities of the AM process chain is presented in Section 2. A case study of an AM cyber-physical attack, in which the .STL file structure is altered, is presented in Section 3. The resulting effectiveness of this attack is reported in Section 4 via part testing and experimentation with human subjects. Finally, in Section 5, the results of the attack are analyzed to identify ways of preventing and mitigating future attacks.