A novel Physical Unclonable Function using RTN

Abstract

PUFs have emerged as an alternative to traditional Non-Volatile Memories in the field of hardware security. In this paper, a novel PUF is proposed that uses the Random Telegraph Noise phenomenon as the underlying source of entropy. This phenomenon manifests as discrete and random shifts in the drain current of transistors and it is characterized by several parameters like the number of the defects in the device, as well as the emission and capture time constants and current shifts of these defects. Using the recently reported Maximum Current Fluctuation metric, it is possible to condense all this information and use it for the PUF design. By forming pairs of transistors, measuring, and comparing their Maximum Current Fluctuation over a given time interval, we demonstrate using numerical experiments that it is possible to obtain a PUF. Furthermore, the results reported here show that this RTNbased PUF matches, and even outperforms, other silicon PUFs in terms of uniqueness, unpredictability, and reliability with an evident advantage in silicon area.

High-level design of a novel PUF based on RTN

“…Unfortunately, many silicon PUFs lack sufficient robustness against aging, since the threshold voltage is critically impacted by phenomena like BTI or HCI [8], thus degrading all quality metrics. Recently, a solution has been presented that uses the Random Telegraph Noise (RTN) [9] as an entropy source together with the Maximum Current Fluctuation (MCF) metric [10] to attain an aging resilient PUF [11]. RTN is a time-dependent variability phenomenon observed as discrete and random shifts of the threshold voltage that translates into drain current shifts.…”

“…In this way, the MCF value is determined by all characteristic parameters of RTN defects (the emission/capture time, the current shifts, and the number of defects) [10]. Thus, as demonstrated in [11], comparing the MCF of two transistors allows to obtain a reliable PUF response using RTN. The MCF is insensitive to mismatchinduced variations in the mean values of the drain currents, so the only entropy source used is RTN.…”

“…In essence, this PUF's response bit comes from comparing the MCF of two identical transistors. While the viability and quality of this solution were demonstrated in [11], the goal of this paper is to approach its high-level design, considering a set of non-idealities of the building blocks to carry out the specification refinement process. Using a combination of high-level simulators and description languages together with an RTN simulation tool [13], the quality of the PUF will be assessed and maximum allowable values of the different non-idealities will be extracted so that a transistor-level design can be carried out.…”

On the use of an RTN simulator to explore the quality trade-offs of a novel RTN-based PUF

“…Typically, silicon PUFs use TZV as the entropy source [8]. However, new solutions have emerged that can use TDV instead: in [9] and [10], RTN is used as the entropy source for this. One salient advantage is that RTN is a mechanism that can be less sensitive to circuit aging provided that the biasing of the transistors and their geometries in the PUF are adequately set.…”

“…RTN is observed as random and sudden discrete shifts of the drain current, caused by threshold voltage shifts associated with stochastic charge trapping/de-trapping events in/from device defects or traps [2]. A PUF based on this mechanism takes two identically designed, equally biased elements (e.g., 2 ring oscillators or 2 transistors in [9] and [10], respectively) and generates a digital response bit (i.e., "0" or "1"), according to which of the elements "contains" more RTN (e.g., how many deviations from the nominal oscillation frequency, caused by RTN, each ring oscillator displays). This response is the basis for the authentication/identification protocol.…”