New breakthrough in ‘digital fingerprint’ tech dramatically boosts smart device security
The unclonable ‘digital fingerprint’ technology resists attacks from advanced AI-powered systems.
CUHK engineers have created a new type of unclonable ‘digital fingerprint’ that strengthens security in smart devices and resists attacks from advanced AI-powered systems. The innovative solution leverages nanomaterials to build microchips with unique, reconfigurable identifiers, promising stronger cybersecurity for autonomous vehicles, robotics, drones, and Internet of Things (IoT) networks. The research findings have been published in Nature Communications.
As more devices process data locally rather than through cloud servers, security risks such as hacking, physical cloning, and reverse engineering are rising. Hardware-based authentication tools can make devices harder to clone, but existing technologies are becoming increasingly vulnerable to modern AI-driven cyber threats.
Stronger protection against emerging threats
To address this, a team led by Prof. Guohua Hu, Dr. Yang Liu, and Dr. Jingfang Pei of CUHK Department of Electronic Engineering has developed an enhanced version of a common hardware authentication tool: the physical unclonable function (PUF).
Every microchip has tiny, uncontrollable physical variations that occur naturally during the manufacturing process. A PUF turns those physical quirks into a unique ‘fingerprint’ by sending a specific electronic signal pattern that generates a unique ‘response’ from the chip. This response acts like a built-in password that the chip can produce on demand, making it a powerful tool for authentication and secure communication.
However, traditional silicon-based PUFs cannot be reconfigured, meaning their unique patterns remain the same throughout the device’s life. This makes them vulnerable to modern AI and machine-learning attacks that can learn and predict their responses over time.
The CUHK team has developed a more secure alternative by using carbon nanotubes instead of silicon. Carbon nanotubes are one-dimensional nanomaterials that transmit electrical signals with high efficiency and can be customised for different electronic applications. During manufacturing, they scatter randomly across a chip, creating a unique physical pattern that cannot be cloned. Unlike conventional approaches, this pattern can also be updated and reconfigured over time to stay ahead of attackers.
Carbon nanotubes deliver unprecedented reconfigurability
Using carbon nanotubes, the team developed PUFs that can be programmed into more than 10 trillion possible configurations. The prototype devices demonstrated strong randomness, uniqueness, and reliability, outperforming earlier solutions.
The team develops PUFs that can be programmed into more than 10 trillion possible configurations.
The team also tested the PUFs against advanced AI and machine learning attacks. Experimental results showed that these attacks achieved only about a 50–60% success rate in predicting the PUF responses, essentially no better than random guessing and far too low to pose any threat.
To showcase real-world potential, the researchers built an authentication protocol for self-driving vehicles and tested it by simulating traffic in Hong Kong’s Central district. The system supported fast authentication with minimal latency and low processing demands. Looking ahead, the team aims to adapt its fabrication approach for industrial use and integrate the technology into standard chip manufacturing processes, while adding additional components and functions to make the system fully deployable across smart devices, autonomous systems, and IoT applications.
Read more: