SUTD researchers developed phase-change key for new hardware security
SUTD - Shao-Xiang Go, Qiang Wang, Kian Guan Lim, Tae Hoon Lee, Natasa Bajalovic and Desmond K. Loke
As more and more data are being shared and stored digitally, the number of data breaches taking place around the world is on the rise. Scientists are exploring new ways to secure and protect data from increasingly sophisticated cyber attacks. A new type of reconfigurable, scalable, low power hardware security device with high resilience to AI attacks has been developed by researchers using phase-change materials. This research is led by Assistant Professor Desmond Loke from the Singapore University of Technology and Design (SUTD).
Dr Loke said, “We developed a novel hardware security device that can eventually be implemented to protect data across sectors and industries, as breaches in private data have been ever-increasing.”
The device, known as the physical unclonable function (PUF), is a new type of phase-change PUF that is scalable, more energy-efficient and secure against AI attacks compared to traditional silicon PUFs. This is due to the electrical and physical properties of phase-change materials, along with the fabrication process.
Amorphous-based PUF (A PUF). a) Schematic illustration of the product authentication concept. The goods/packages can be attached with the A PUF by the original manufacturer. A unique PUF ID will be generated for each product. By accessing enrolled keys in a secure cloud database, end users, for example, consumers and other users, can validate the product and certify the source. Other useful information such as the product expiration date, supply chain path and other information relevant to the targeted product can also be stored in the A PUF. b) New types of PUF can be designed using phase-change materials (PCM), based on the reversible switching between the melt-quenched amorphous state and fully crystallized state of a chalcogenide layer, along with sensing different as-deposited amorphous states, showing a large contrast in electrical and optical signatures. In this work, unique cryptographic keys are generated by the large degree of randomness of the as-fabricated/melt-quenched amorphous state of an ensemble of A PUF cells. The A PUF is implemented based on the design flow: i) the cell resistance is read using a differential readout strategy, ii) high stochasticity of the as-fabricated or melt-quenched state is enabled by different fabrication/cycling conditions, iii) the stochasticity is reflected by the difference in the resistance of cells, and iv) the difference in the as-fabricated/melt quenched state is utilized to generate cryptographic keys.
The SUTD research team fabricated a group of phase-change devices, which switch reversibly between the glassy amorphous state and crystal orderly state. They then used the variation in the device’s electrical conductance to construct the PUF due to the inherent randomness arising from the manufacturing process, which is not shown by conventional silicon-based devices.
The researchers modelled the characteristics of actual phase-change devices to generate a simulation of many more phase-change-based PUFs. Dr Loke and his team used machine learning, a method that allows AI to study a system and find new patterns, to test the PUF’s security. To examine if the AI could use this training to make predictions about the encrypted key and reveal system insecurities, the researchers trained the AI with the phase-change PUF simulation data.
Dr Loke added, “Normal humans are not able to develop a model from a vast amount of data, but neural networks could. We also found that it was not possible for the encryption process to be learned and that the AI could not develop a model to decrypt the phase-change PUF.”
As potential hackers could not use the ‘stolen’ key to reverse engineer a device for future use, the resistance to machine learning attacks makes the PUF more secure. The phase-change PUF can also create a new key immediately through the reconfiguration mode if the key is hacked.
The phase-change PUF could be used in a variety of applications with these features, together with the capacity to operate at high temperatures. Future research can pave a way for its use in household devices, printable and flexible electronics and other devices.
The co-authors include Shao-Xiang Go, Qiang Wang, Kian Guan Lim, Natasa Bajalovic from SUTD, and Tae Hoon Lee from University of Cambridge.
Acknowledgements:
The authors thank T. C. Chong (SUTD), R. Zhao (Tsinghua University), S. R. Elliott (University of Cambridge), and F. A. Spaepen (Harvard University) for valuable discussions. This work was financially supported by the Ministry of Education (Singapore) (MOE-T2EP50220-0022), SUTD-MIT International Design Center (Singapore), SUTD-ZJU IDEA Grant Program (SUTD-ZJU (VP) 201 903), SUTD Kickstarter Initiative (SKI 2021_02_03, SKI 2021_02_17), Changi General Hospital (Singapore) (CGH-SUTD-HTIF2019-001), Agency of Science, Technology and Research (Singapore) (A20G9b0135) and National Supercomputing Centre (Singapore) (15 001 618). S.X.G. thanks the Singapore University of Technology and Design for scholarship support.
Reference:
Ultrafast Near-Ideal Phase-Change Memristive Physical Unclonable Functions Driven by Amorphous State Variations, Advanced Science. (DOI: 10.1002/advs.202204453)