Carnegie Mellon University

laptop on a black background

August 03, 2023

Detecting Trojans in Analog Hardware

By Hope Reveche

Krista Burns

Globalization has allowed for better collaboration in both the academic and industry realms; however, it has also led to greater risks for security breaches. A large threat in the world of cybersecurity is hardware trojans, malicious alterations that disrupt an integrated circuit, the foundation for all modern electronics. Using trojans, attackers can control devices, steal personal data and information, or deploy unwanted software.

Digital and analog technologies differ in the way they transmit information. Whereas the digital domain discretizes information into 0s and 1s, the continuous analog domain allows for infinite values for measuring, making them more complex to defend against hardware trojans. As a result, very little research has been done in this analog space.

Researchers in electrical and computer engineering saw this gap in research and designed a hardware trojan detection method meant to continuously monitor for malicious hardware. The study involved creating two novel hardware trojans to test with their electromagnetic (EM) side channel detection sensor that helps with trojan detection and classification.

The sensor is placed so that it can monitor power lines that are impacted by the trojan. A signal then gets compressed and expanded so that its pattern can be generated over and over again, ready to be matched to any incoming signals. The current issue is when the sensor doesn’t recognize a certain signal – when it comes to hardware trojans, this can be a great risk. “It’s like this,” says John Kan, Ph.D. student in electrical and computer engineering and main author of the paper. “You train a system and tell it A, B, C, D, E. But, then you give it F and it has no idea what to do. It may call it E, it may call it D, it may call it B. It could get it wrong every time.”

The new method allows for these unknown signals to be recognized and learned. When the system’s decoder attempts to match received signals with what is already in its database, it has the ability to alert the user that there is an issue if the signals don’t match. The team tested their hypothesis with great success and hopes to develop the method further to be able to fill all gaps in the detectable patterns.

“We're protecting things that rely on analog circuitry,” Kan says. “So, for example, autonomous vehicles, which have radar. The car senses what's in front of me, and then it'll bounce back and give me a distance estimation. If somebody were to hide something in the circuit though, they could cause an outage, or make something appear that wasn't there, causing the car to stop real fast, and that could be really catastrophic.” 

Moving forward, the research team hopes to optimize the sensor’s build, shrinking it down to a more compact size. Once this is achieved, the goal is to implement it into everyday technologies like cars and planes, making analog circuits less susceptible to hardware trojans.

The team of researchers included John Kan; Yuyi Shen; Jiachen Xu; Ethan Chen; Bo Yuan Yang; Jimmy Zhu, professor of electrical and computer engineering; and Vanessa Chen, assistant professor of electrical and computer engineering. Their findings were recently published in Sensors and the Institute of Electrical and Electronics Engineers Open Journal of Circuits and Systems.