How To Crash A Drone By Hacking Its 3D Propeller Design
Researchers from Israel's Ben-Gurion University of the Negev and two other universities show how attackers can exploit 3D manufacturing processes.
October 20, 2016
Researchers at Israel's Ben-Gurion University of the Negev along with their counterparts at the University of South Alabama and the Singapore University of Technology and Design have demonstrated how attackers can cause fatal equipment failures by sabotaging the integrity of 3D-printed parts.
In a proof-of-concept exploit described in a technical paper and video released this week, the researchers crashed a $1,000 quadcopter by manipulating the design of a 3D-printed replacement propeller used by the drone.
For the demonstration, the researchers simulated how an attacker might use a standard email phishing lure to gain access to the computer used to control the 3D printing of replacement drone-propellers - and drop malware on it.
The malware would let the threat actor break into the system, search for the blueprint for the propeller, download and alter it, and then swap out the original design file with the doctored one.
For the proof-of-concept exploit, the researchers inserted sub millimeter-scale internal cavities - or voids - in the design of the propeller so it would deteriorate more quickly than normal under operational conditions yet appear intact when inspected visually.
When the weakened propeller is then used in the drone, it fails quickly, causing the aircraft to crash.
The exploit is designed to expose the threats manufacturing companies face from the increasing use of additive printing, aka 3D printing, for functional parts in manufactured equipment, the researchers said in a statement.
"With the growth of additive manufacturing worldwide, we believe the ability to conduct malicious sabotage of these systems will attract the attention of many adversaries, ranging from criminal gangs to state actors," said Yuval Elovici, a member of the team that developed the exploit.
More than 100 industries, including aerospace, defense, and automotive currently use 3D-printed parts to varying degrees in their products, the statement noted.
The new exploit is not the first to demonstrate the vulnerability of 3D-printed parts to such tampering. Other researchers have already discussed the possibility of modifying the design files for 3D printers and of modifying manufacturing parameters like the temperature of the extruder in certain types of 3D printers, notes Mark Yampolskiy, assistant professor at the University of South Alabama’s School of Computing.
But prior research has focused mainly on the reduction of mechanical properties, like the tensile strength of a 3D manufactured object, so it would fail when put into operation.
"Our proposal introduces defects that increase fatigue," Yampolskiy says. "This means that parts will not immediately break under operational conditions - and can pass mechanical tests."
The proof-of-concept shows how an attacker can introduce a defect that causes fatigue to develop faster in a functional part, causing it to break, he says.
The technical paper is unique in that it presents a full chain of attack on an additive manufacturing process. It starts with a cyberattack on the system with the design files and continues through malicious modification of the file, and manufacture of the compromised part, he says.
And unlike previous attacks that have targeted specific printer models, this research is the first one that is close to universally applicable to any 3D printer, according to the reserachers.
The likelihood of this type of attack depends on a variety of factors. Typically, a professional 3D printer used for additive manufacturing is unlikely to be directly connected to the Internet, Yampolskiy says. So while direct attacks on such systems might be hard to pull off, the systems can still be accessed indirectly via other computers with which they might be connected.
"In this paper, we have tried to analyze whether an adversary with a limited sophistication can perform such an attack. And the answer is – as scary as it is – yes," he says.
Related stories
About the Author
You May Also Like
Cybersecurity Day: How to Automate Security Analytics with AI and ML
Dec 17, 2024The Dirt on ROT Data
Dec 18, 2024