CoChat

Robust Authentication of Consumables With Extrinsic Tags and Chemical Fingerprinting

Open access

The DOI matches its Crossref record; 19 of the 19 references it cites were found in a registry or an index; none of the 30 sources its claims cite is checked for support yet.

Abstract

Consumables-from food to pharmaceuticals and supplements-are becoming increasingly vulnerable to various modes of counterfeiting due to the growing complexity of their supply chain. Mislabeling, re-branding, and false advertising are prevalent in this sector. Existing physical authentication techniques fail to adequately verify integrity of these products and protect the end-users. In this paper, we aim at addressing this critical problem through the development of a novel authentication solution. It builds on the chemical analysis properties of a powerful spectroscopy technique, namely, nuclear quadrupole resonance (NQR) that is quantitative, non-invasive, low-cost, and amenable for miniaturization (to hand-held form factors). The method is sensitive to small variations in the solid-state chemical structure of a sample, which change the NQR signal properties. These attributes can be unique for various manufacturers, enabling their use as manufacturer-specific watermarks. However, NQR spectroscopy only works reliably (i.e., provides good sensitivity) on compounds that contain certain nuclear isotopes. We take advantage of the intrinsic properties of NQR-sensitive isotopes to use them as extrinsic tags in NQR-insensitive products. The NQR spectra of these extrinsic tags act as unique watermarks that can be analyzed using machine learning methods to authenticate any consumable with high confidence. In particular, we use support vector machines to classify the measured spectra and confirm the identity of items under test. We have assessed this approach on a variety of consumables utilizing semi-custom equipment and verified that it results in high (>95%) classification accuracy. In order to prove the unclonability of such extrinsic tags, we have also performed a mathematical analysis that proves the randomness of the extrinsic tag and confirms its robustness to brute-force attacks.

Overview

CoChat Summary

TL;DR

The paper proposes nuclear quadrupole resonance (NQR) spectroscopy with intrinsic and extrinsic chemical tags plus SVM-ECOC classification to authenticate consumables and trace them through supply chains. It reports >95% classification accuracy in proof-of-concept tests and a randomness-function estimate of 2.55×10^161 for a three-component extrinsic tag.

Key findings

  • NQR parameters—amplitude (A), effective decay time (T2,eff), and linewidth (f)—serve as chemical fingerprints; SVM-ECOC classification gave >95% accuracy, with cross-validation accuracy of 97.5% for one correctly labeled dataset versus a permutation null distribution of 42% ± 3.35% (16σ separation).
  • Extrinsic tags were demonstrated: L-histidine in common sugar at 30% and 40% mass was classified with 100% cross-validation accuracy; multiple tags (L-histidine and L-proline) in milk powder were distinguished; a sodium chlorate 35Cl tag was detected at 0.1 g within ~40 sec, ~20× better than an L-histidine 14N tag, whose sensitivity was ~5 cm3 (~2.1 g powder) within 40 sec.
  • Manufacturing and physical form affected NQR linewidth: custom L-histidine pills versus powder showed ~10σ linewidth change, and Tylenol caplets versus gelcaps were classified with 97% cross-validation accuracy.
  • A randomness function for three compounds with five NQR-sensitive atoms each, using assumed parameter ranges and sensitivities, yielded Ψ = 2.55×10^161, supporting unclonability against brute-force attacks.

Methods

  • NQR spectroscopy used the spin-locked spin echo (SLSE) pulse sequence; matched filtering and nonlinear fitting extracted A, f, and T2,eff, which were used as features for SVM with error-correcting output codes (ECOC) for multi-class manufacturer classification.
  • The experimental setup comprised an inductive solenoid coil and programmable impedance matching network inside a Faraday cage, connected to a commercial bench-top MR spectrometer (Kea2, Magritek), powered by two 12 V, 18 Ah lead-acid batteries and controlled via the Prospa GUI; samples included acetaminophen, L-histidine, L-proline, sugar, milk powder, and sodium chlorate.
  • Statistical validation used a bootstrapping-like synthetic-noise procedure for parameter distributions and permutation testing (10,000 iterations); computational complexity of SVM training and prediction was also estimated.

Limitations

  • Stated: the current setup is a bench-top device too large and expensive for average consumers; future work aims to miniaturize with a system-on-module, improve sensitivity with pre-polarization, and move computation to the cloud. Current applications are limited to consumables, with future work on non-consumables.
  • Stated: NQR is relatively insensitive, especially 14N at low frequencies (0.1–5 MHz), requiring sufficient tagging material and measurement time; environmental RFI and temperature fluctuations require shielding and control. Extrinsic tags must be safe to consume, nonreactive, and not alter product properties. The current setup is limited to solid tags because NQR is a solid-state technique, with plans to combine NQR with low-field NMR for liquids.
  • Evident from the text: proof-of-concept experiments used small training sets (maximum 40 instances and 3 features), some analyses assume mono-exponential decays, and re-substitution accuracy is positively biased, making cross-validation more suitable; measured linewidth is convolved with the acquisition window, adding a constant offset.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.1109/access.2019.2893518

Cites

19 of 19 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Verification

Claims the paper makes

0 Valid · 12 Not valid · 17 Not checked · 87 Own results and interpretations

53 more claims

Create a free account to see all 116 claims and every source verdict.

See all claims — free account
  • It may state two separate claims as one. 31
  • Probably not a substantial claim. 37

Create a free account to see these claims.

Citations

Cites

Research works printed but not linked: 10

  1. 10.1007/s00723-012-0320-2 · DOI
  2. 2017, Global Dietary Market is Expected to Reach Around USD 220 3 Billion in 2022
  3. 0, Direct food substances affirmed as generally recognized as safe
  4. 2017, 1 in 10 Medical Products in Developing Countries is Substandard or Falsified
  5. 2019, What You Need to Know
  6. 2016, Largest-Ever Seizures of Fake Food and Drink in Interpol-Europol Operation
  7. jahnke, 2017, Bad Medicine
  8. latosinska, 1995, NQR spectroscopy in definition of biological activity of medicines and pesticides, Mater Presented a All-Polish Seminar Nucl Magn Reson Appl, 29, 134
  9. 2016, The Fight
  10. leeper, 2019, Permutation Tests

Cited by

See all 15

Metadata

Authors
Masna, N. V. R., Chen, C., Mandal, S., & Bhunia, S.
Affiliations
  • University of Florida (US)
  • Case Western Reserve University (US)
Venue
IEEE Access
Published
2019
Volume
7
Pages
14396-14409
ISSN
2169-3536 (ISSN-L)
OpenAlex
W2908776200
Cited by
15
References
19
Citations by year
2024: 2 · 2023: 4 · 2022: 3 · 2021: 4 · 2020: 1 · 2019: 1
FWCI
1.87
Type
journal article
Language
en
Open access
Locations
  • IEEE Access (publishedVersion, other-oa)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
Funders
National Science Foundation (SHF-1563688, SHF-1563924)
Topics
  • Physical Sciences › Computer Science › Hardware and Architecture › Physical Unclonable Functions (PUFs) and Hardware Security
  • Physical Sciences › Chemistry › Analytical Chemistry › Spectroscopy and Chemometric Analyses
  • Physical Sciences › Computer Science › Information Systems › QR Code Applications and Technologies
Keywords
counterfeiting, chemical fingerprinting, support vector machine, spectral classification

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

A language model drew the claims from the paper on 2026-11-19, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).

A second model labelled each claim on 2026-11-19 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.