The proliferation of the internet of things (IoT) and its global interconnectivity has led to an exponential increase of digital information across every sector of modern society. Each day an enormous amount of data is generated and transferred across the world in numerous sectors, such as defense, health, automation, etc. With this increasing demand for networked smart objects, there is an equally growing demand for ensuring the security and reliability of these objects. Acknowledging the limitations associated with conventional software cryptographic primitives, a vast amount of interest has been directed towards improving hardware security.

The field of hardware security focuses on providing cryptographic primitives on hardware and protecting chips from adversarial attacks. For over a decade, the physically unclonable function (PUF) has remained a promising approach towards achieving reliable primitives in hardware security. PUFs exploit natural variations in the physical microstructures of hardware components and their complex interactions with different stimuli (voltage, magnetic field, light, etc). Therefore, PUFs are considered akin to chip fingerprints that can be used in a wide range of security applications, such as identification, authentication, and cryptographic key generation. Traditional PUFs which rely on silicon are steadily declining in advancement and are highly vulnerable to machine learning and new attack models, prompting a need for PUFs based on novel materials and silicon alternatives.

Ever since the discovery of graphene, graphene-based electronics have revolutionized the electronics community due to their intriguing physical and chemical properties. However, commercial large-scale manufacturing has yet to produce defect-free graphene that can achieve high electrical performance consistently. We harness these inherent disorders associated with carrier transport in graphene field-effect transistors (GFETs) grown using chemical vapor deposition (CVD) to generate a graphene-based PUF.

We show that the Dirac voltage, Dirac conductance, and carrier mobility of a large population of GFETs follow a Gaussian random distribution, allowing the GFETs to be used as a PUF. Furthermore, we employ various statistical measures such as hamming distance, correlation coefficient, and entropy to measure the robustness and strength of the GFET PUF. The strength of the GFET PUF is further tested against machine learning models such as the predictive regression model and generative adversarial network (GAN), and was observed to be extremely resilient to advanced modeling attacks. In addition, reconfigurability, a major bottleneck of current silicon technology, is addressed by the graphene-based PUF through the exploitation of the memristive properties of graphene. This indicates graphene-based PUFs as a viable solution for next-generation secure electronics. We also measure the consistency of the GFET PUF over extreme conditions of temperature, supply voltage, and time, demonstrating that these primitives are extremely reliable. Finally, the graphene-based PUF is shown to be scalable and extremely energy efficient, checking the crucial components of IoT and thereby promoting solutions to a sustainable future.

 To know more about the work, please refer to the paper “Graphene-based physically unclonable functions that are reconfigurable and resilient to machine learning attacks” published in Nature Electronics following the link: https://www.nature.com/articles/s41928-021-00569-x