Date of Graduation


Document Type


Degree Type



Eberly College of Arts and Sciences


Forensic and Investigative Science

Committee Chair

Luis Arroyo

Committee Member

Tatiana Trejos

Committee Member

Marina Galvez


Counterfeit pharmaceuticals are an actively developing health and economic threat worldwide. Particularly prevalent are counterfeit pharmaceuticals distributed in emerging nations and through internet pharmacies or e-pharmacies. Although technology has been developed that discourages anti-counterfeiting practices (such as optically variable devices, invisible ink, and track-and-trace technology), it remains somewhat novel and expensive to implement on a widespread scale.

In this study, Laser Induced Breakdown Spectroscopy (LIBS) and Attenuated Total Reflectance-Fourier Transform Infrared Spectroscopy (ATR-FTIR) were proposed as fast and non-invasive tools for the identification of counterfeit pharmaceutical packages. The main objective of this research was to develop and evaluate the capabilities of LIBS and ATR-FTIR to determine chemical differences between counterfeit and authentic pharmaceutical packaging samples. LIBS and ATR-FTIR possess several characteristics that render them suitable for rapid on-site detection. They produce analytical results in less than one minute per sample, with high sensitivity and selectivity, limited sample preparation, and minimal destructivity.

The methods were evaluated through the analysis of a dataset of 166 packages (112 counterfeits and 54 authentic sources). The dataset was divided into two main subsets. The first subset was evaluated to identify the informative value of LIBS for fast screening of black barcodes and the carton substrate (100 counterfeit and 35 authentic). The multi-color inks and paper of the second subset was investigated for variation of chemical profiles within and between sources, and the method’s capabilities to distinguish between counterfeits (112) and authentic samples (12).

One hundred and twelve counterfeit pharmaceutical cartons were printed from five different sources, mimicking six authentic counterparts. The authentic subset consisted of twelve secondary packages of six common medical products, including packages from the same and different manufacturing lots. The selected products consisted of vasodilators, antivirals, steroids, and other commonly counterfeited pharmaceuticals.

Intra-source variation of the counterfeit subset was investigated; it was determined to be sufficiently lower than inter-source variation. False exclusion rates were calculated to be less than 20% for samples originating from the same source (e.g., same package, intra-lots, replicate printouts).

Using LIBS, a two-class classification system was used for the combined black barcode ink and paperboard carton spectra (n = 135, 100 counterfeit, 35 authentic packages). As black barcode ink is very common on pharmaceutical packaging, this system was used as a general screening technique to quickly identify a sample as authentic or counterfeit, regardless of counterfeit printing source. In general, the correct classification rates for this set were over 92%.

The classification models were established using six machine learning methods: Random Forest, Naïve Bayes, Neural Networks, k-Nearest Neighbors, Quadratic Discriminant Analysis, and Linear Discriminant Analysis. A random split of 60% and 40% of the dataset was applied for training and testing of the classifier algorithms. Principal Component Analysis (PCA) was utilized on the LIBS and ATR-FTIR data for variable reduction purposes. The principal components for each ink type were combined prior to classification.

Also, a six-class system was also used to classify the dataset using LIBS, ATR-FTIR, and combined data from both techniques (n = 124, 112 counterfeit, 12 authentic packages). The machine learning methods classified the samples as belonging to one of five counterfeit printing sources or their corresponding authentic counterpart. Seven ink colors (red, blue, yellow, green, brown, pink, black) were analyzed; additionally, in ATR-FTIR, the paperboard substrate was also analyzed. In most comparisons, LIBS had a successful classification rate of over 70% and ATR-FTIR had a successful classification rate of over 85%. When the data from both techniques were combined, the discrimination power of the system increased to 93% correct classification. Although LIBS and ATR-FTIR had a low misclassification rate when used in isolation, the misclassification rate could be reduced even further through data combination.

The results of this study are encouraging for the inclusion of LIBS and ATR-FTIR as a screening method for the detection of counterfeit pharmaceutical packaging. The utilization of combined data to discover chemical signatures addresses an urgent need in the investigation of counterfeit pharmaceuticals. Also, the classification of counterfeit samples into their specific counterfeit source may benefit investigators as they make determinations in the counterfeit pharmaceutical packaging supply chain. This study is anticipated to offer relevant tools to both government and pharmaceutical industry in the detection and fight against counterfeit pharmaceuticals.