Edible unclonable functions

Abstract

Counterfeit medicines are a fundamental security problem. Counterfeiting medication poses a tremendous threat to patient safety, public health, and the economy in developed and less developed countries. Current solutions are often vulnerable due to the limited security levels. We propose that the highest protection against counterfeit medicines would be a combination of a physically unclonable function (PUF) with on-dose authentication. A PUF can provide a digital fingerprint with multiple pairs of input challenges and output responses. On-dose authentication can verify every individual pill without removing the identification tag. Here, we report on-dose PUFs that can be directly attached onto the surface of medicines, be swallowed, and digested. Fluorescent proteins and silk proteins serve as edible photonic biomaterials and the photoluminescent properties provide parametric support of challenge-response pairs. Such edible cryptographic primitives can play an important role in pharmaceutical anti-counterfeiting and other security applications requiring immediate destruction or vanishing features.

Fig. 1. Combination of PUF and on-dose authentication for anti-counterfeiting of medicines.

a Schematic illustration of an on-dose PUF with a photograph of covert and transparent PUFs attached on the surface of medicines. The PUF device is composed of nothing but proteins from fluorescent proteins and silk to be edible and digestible. The distinct photoluminescent properties of fluorescent proteins in silk provide the parametric support of unique challenge-response pairs. In reaction to an input challenge (C_n), the edible PUF generates its corresponding output response (R_n), resulting in a cryptographic key (K_n). The protein-based PUFs attached on the surface of medicines can be used for on-dose authentication of each individual medicine. b Concept of on-dose authentication. Each individual medicine in a solid oral dosage form (e.g., tablets and capsules) is integrated with an edible PUF device by the pharmaceutical manufacturer. End users (e.g., pharmacists and consumers) can ensure the provenance and validate the medicine by accessing the enrolled digital keys in a secure database (e.g., cloud server). In addition, this edible PUF could be utilized to provide dose information and manufacturer-determined data, including product information (e.g., dosage strength, dose frequency, and expiration date), manufacturing details (e.g., location, date, batch, and lot number), and distribution path (e.g., country, distributor, wholesaler, and chain).

Fig. 2. All protein-based edible PUFs made of silk and fluorescent proteins.

a Regenerated particulate eCFP, eGFP, eYFP and mKate2 silk produced by silkworm transgenesis via the piggyBac transposase method in which fluorescent proteins and silk (i.e., Bombyx mori) are genetically fused as recombinant proteins. The fluorescent images are acquired with a set of the excitation (λ_ex) and emission (λ_em) wavelengths, as specified on top of each photograph. The scale bar is 5 mm. b Photograph of an edible PUF device in which fluorescent silk microparticles are embedded in a thin silk film. The scale bar is 2 mm. Scanning electron microscopy (SEM) image of fluorescent silk microparticles with zeolite-like shapes. The scale bar is 200 μm. The inset shows a higher magnification SEM image of fluorescent silk microparticles. The scale bar is 50 μm. c The emission spectra of particulate eCFP (cyan solid line), eGFP (green solid line), eYFP (yellow solid line), and mKate2 (red solid line) silk cover a relatively broad wavelength range in the visible light, while the emission peak positions are not overlapped among others. d Confocal fluorescence microscopy images of the corresponding fluorescent silk microparticles under excitation wavelengths of 405, 458, 514, and 561 nm, respectively. The scale bar is 100 μm. The size of fluorescent silk microparticles is 99.3 ± 7.9 μm (mean ± standard deviation) (Supplementary Fig. 3).

Fig. 3. Flowchart of cryptographic key generation from an edible PUF device.

Randomly distributed fluorescent silk microparticles provide an entropy source. An input challenge (C_n) is determined by a set of the excitation and emission bands, which is optimized for each type of fluorescent silk (i.e., eCFP, eGFP, eYFP, and mKate2 silk). The corresponding fluorescent image acts as an output response (R_n). A resultant digitized key (K_n) is generated by undergoing four steps in the extractor. The extractor is designed to generate uniform and reproducible digital keys after data acquisitions of fluorescent images. For each response returns in a 64-bit binary key, a total key size of this PUF is 256 bits. The key size can further be enhanced by an increase in the density of fluorescent silk particles in the PUF device or by additional combinations of excitation and emission bands (Supplementary Fig. 11).

Fig. 4. Basic elements in the extractor.

a The reading apparatus acquires raw fluorescent images of responses (R_n) of an edible PUF device in which an admixture of eCFP, eGFP, eYFP, and mKate2 silk microparticles is embedded in a thin silk film (300 pixels × 300 pixels). b The peak finding and binarization processes provide high stability and reproducibility in key generation, reducing the number of pixels to 50 pixels × 50 pixels. c The von Neumann debiasing process allows to compress dominant 0-bits resulting from the relatively small number of peaks. In simple von Neumann debiasing, the rate of compression is too high such that the raw data size needs to be much larger than an extracted size. In our extractor, the two-pass tuple-output von Neumann debiasing algorithm maintains a practical data size. After von Neumann debiasing, first 64 bits in each digitized key (K_n) are selected to create a total of 256-bit security key.

Fig. 5. Representative binary bitmap of cryptographic keys.

Digitized security keys (256 bits in each PUF) are generated from 30 different edible PUF devices. Notably, the bitstream from this binary bitmap is random, validated by the NIST randomness tests (Table 1 and Supplementary Table 2). The production and fabrication of edible PUF devices are easily scalable due to the choice of materials and the fabrication strategy.

Fig. 6. Characterizations of the basic performance matrices of edible PUFs.

a Bit uniformity calculated from 30 different PUFs shows the unbiased distribution of 0 and 1 states after von Neumann debiasing with a mean (μ) value of 0.5. b Inter-device Hamming Distance (HD) is used to characterize the device uniqueness of 30 different edible PUFs. A Gaussian fit of the histogram returns μ = 0.5032 and a standard deviation (SD; σ) of 0.0458 where the probability density of Gaussian distribution $P\left(x\right)=\frac{1}{\sigma \sqrt{2\pi }}{\mathrm{e}}^{-{\left(x-\mu \right)}^2∕\left(2{\sigma }^2\right)}$. The mean value is close to the ideal value of 0.5 with a narrow distribution. The mutually independent bits (degree of freedom or number of independent variables) of 120 (≈0.5032 × (1 − 0.5032)/0.0458²) result in an encoding capacity of 2¹²⁰ (≈1.3292 × 10³⁶). Intra-device HD is used to characterize the readout reproducibility (i.e., bit error rate) from 10 repeated challenge-response cycles (nine pairwise comparisons) in each PUF device among 30 different edible PUFs. A Gaussian fit of the histogram returns μ = 0.0632 and σ = 0.0164. The clear separation between the intra-device and inter-device distributions indicates extremely low false positive and negative rates (Supplementary Fig. 14). c Comparisons among digitized keys from four responses in each PUF ensure the uniqueness of the responses in the PUF device. A mean HD is calculated in each PUF and 30 mean HDs from 30 different PUFs are plotted with μ = 0.4990 and σ = 0.0041. d Pairwise comparisons among 30 different edible PUFs. The HDs in the off-diagonal areas fluctuate near the mean value of 0.5032. The source data are provided as a Source Data file.