Improving the longevity of optically-read quantum dot physical unclonable functions

Abstract

Quantum dot physically unclonable functions (QD-PUFs) provide a promising solution to the issue of counterfeiting. When quantum dots are deposited on a surface to create a token, they form a unique pattern that is unlikely to ever be reproduced in another token that is manufactured using the same process. It would also be an extreme engineering challenge to deterministically place quantum dots to create a forgery of a specific device. The degradation of the optical response of quantum dots over time, however, places a limitation on their practical usefulness. Here we report methods to minimise the degradation of photoluminescence (PL) from InP/ZnS quantum dots suspended in a polymer and demonstrate reliable authentication using a fingerprinting technique to extract a signature from PL, even after significant degradation has occurred. Using these techniques, it was found that the addition of a polylauryl methacrylate (PLMA) copolymer improved the longevity of devices. The best performing example of this was the Polystyrene-PLMA based material. From this, it is projected that 1000 bits of information could be extracted and read after a period of years, therefore providing a compelling solution to the issue of counterfeiting.

Figure 1

(a) A schematic of the apparatus used to measure PL intensity maps and to captures images of each token. The quantum dots in each token were excited using white light filtered through a 450 nm short-pass filter. A 500 nm long-pass filter placed in front of the CCD ensures only light from the emission of the QD-PUF is measured. The entire apparatus is sealed within a closed black box when measurements are in progress. (b) Using the apparatus in (a) each token is imaged on the day of its creation and each subsequent day after. The fingerprint generated on the subsequent days is compared to that of day 0 to determine if it matches the original fingerprint.

Figure 2

(a) Top—Images showing photoluminescence (PL) intensity maps from each formulation of the Group 1 tokens captured on the day they were created (day 0). Bottom—PL images of each Group 1 token type captured on day 14. The intensity of day 14 images has been increased by 30% post-capture to aid visual comparison. (i) PMMA, (ii) PMS, (iii) PS, (iv) PVDF, (v) SEBS. (b) The top of each coloured rectangle is the R-LBP generated fingerprint of the labelled QD-PUF on day 0. The middle of each coloured rectangle is the fingerprint from day 14. In the bottom image of each rectangle, the white pixels indicate the pixels that changed in value in the fingerprint between day 1 and 14. The percentage of pixels that changed value and the composition of each is as follows: (i) PMMA, 21.2%, (ii) PMS, 24.5%, (iii) PS, 18.9%, (iv) PVDF, 29.1% (v) SEBS, 19.3%. (c) A plot showing the average PL intensity from each token type as a function of time since their creation. An exponential decay fit was applied to each (dashed line). The grey dotted line represents the background noise signal from the CCD-sensor.

Figure 3

Top—A plot showing the false positive rate (FPR) of each Group 1 token type, which was derived by comparing fingerprints generated from photoluminescence images captured at the times shown after their creation. Bottom—Showing the effective number of independent bits (ENIB) extracted from each Group 1 token as a function of days since their creation.

Figure 4

(a) Top—Images showing the photoluminescence (PL) intensity maps from each formulation of the Group 2 tokens captured on the day they were created (day 0). Bottom—PL images of each Group 1 token type captured on day 14. The intensity of the day 14 image has not been increased post capture, as it was in Fig. 2. For each token type the polymer used was: (vi) PMMAPLMA, (vii) PMSPLMA, (viii) PSPLMA, (ix) PVDFPLMA, (x) SEBSPLMA. (b) The top of each coloured rectangle is the R-LBP generated fingerprint of the labelled QD-PUF on day 0. The middle of each coloured rectangle is the fingerprint from day 14. In the bottom image of each rectangle, the white pixels indicate the pixels that changed in value in the fingerprint between day 1 and 14. The percentage of pixels that changed value and the composition of each is as follows: (vi) PMMAPLMA, 46.5%, (vii) PMSPLMA, 41.2%, (viii) PSPLMA, 19.5%, (ix) PVDFPLMA, 24.12%, (x) SEBSPLMA, 15.4%. (c) The average PL intensity from each token type as a function of time since their creation. An exponential decay fit was applied to each (dashed line). The grey dotted line represents the background noise signal from the CCD sensor.

Figure 5

Top—Showing the false positive rate (FPR) of each token type in Group 2, which was derived by comparing fingerprints generated from photoluminescence captured at the times shown after their creation. Token vi displays no FPR before Day 5 as its value fell below the precision of the implementation of the code used MathWorks MATLAB. Bottom—showing the effective number of independent bits extracted from each Group 2 token as a function of days since their creation.