Artificial Intelligence Approaches for Counterfeit Medication Detection: A Compact Review for Public Health Applications

Review Article Correspondence to: Ivan Arabome, e-mail: [email protected] Copyright: © 2025 The authors. This is an Open Access article distributed under the terms of the Creative Commons Attribution 4.0 International License. How to Cite: Arabome and Arabome (2025). Artificial Intelligence Approaches for Counterfeit Medication Detection: A Compact Review for Public Health Applications. Scholar J Computational Science, 2(11). DOI: 10.5281/zenodo.17506130 Artificial Intelligence Approaches for Counterfeit Medication Detection: A Compact Review for Public Health Applications Ivan Arabome1 and Evan Arabome2 1Department of Data and Computer Science, York St John University, London, UK. 2Department of Community Medicine, Nile University of Nigeria, Abuja, Nigeria. Received 18 August 2025; Acceptance 24 September 2025; Published 2 November 2025. Abstract The global proliferation of counterfeit medicines poses a critical threat to public health, patient safety, and healthcare systems. Traditional detection and traceability mechanisms such as barcoding, serialization, and laboratory analyses are increasingly inadequate against sophisticated counterfeiting methods. Emerging technologies, particularly artificial intelligence (AI), blockchain, and mobile-based systems, have shown promise in supplementing conventional regulatory approaches by enabling proactive, scalable, and data-driven surveillance. This review examines key advances in AI-driven counterfeit detection, with emphasis on visual inspection, machine learning (ML), synthetic data generation, blockchain-based authentication, and mobile or edge deployments. The synthesis identifies how AI enhances counterfeit detection accuracy and speed while highlighting challenges related to data scarcity, algorithmic bias, and ethical oversight. The review concludes that AI and distributed digital verification systems represent transformative tools for pharmaceutical quality assurance, but their impact depends on global data governance, transparent model validation, and equitable access across lowand middle-income settings. Keywords: Artificial Intelligence (AI), Counterfeit Medication Detection, Machine Learning (ML), Blockchain, Synthetic Data, Mobile Health (mHealth), Pharmaceutical Supply Chain, Public Health Surveillance. Scholar J Computational Science 103 Introduction The integrity of the pharmaceutical supply chain is a cornerstone of global public health, ensuring that patients receive safe, effective, and high-quality medicines. However, this integrity is increasingly undermined by the widespread circulation of counterfeit and substandard medical products. According to the World Health Organization (WHO), falsified medicines account for an estimated 10% of all pharmaceuticals in lowand middle-income countries (LMICs), with significant variations across therapeutic classes and distribution channels [1]. These falsified products are not merely economic crimes—they represent a grave public health threat that results in treatment failures, preventable deaths, and the acceleration of antimicrobial resistance, particularly in regions with weak regulatory oversight [2]. The resulting erosion of public trust in health systems further compounds the challenge of delivering effective care. Counterfeit medicines infiltrate supply chains through multiple routes: informal markets, unregulated online pharmacies, and distribution leakages within formal systems. The globalisation of pharmaceutical production and distribution has expanded these vulnerabilities, making traditional border inspections and product recalls increasingly ineffective. This complexity is amplified by digital marketplaces, where consumers can easily access medical products without clear provenance. Consequently, identifying and preventing counterfeit infiltration now requires more than conventional inspection it demands digital traceability, data-driven intelligence, and cross-border collaboration. Traditional approaches to counterfeit detection, such as chemical assays, spectroscopy, chromatography, serialization, and manual packaging inspection, remain foundational to regulatory enforcement. Yet, they are often reactive, time-intensive, and cost-prohibitive. For example, serialization systems while effective in theory require seamless coordination across manufacturers, distributors, and regulators, which remains challenging in fragmented healthcare environments [3]. Likewise, visual inspection by trained personnel is still widely practiced in resource-limited settings but is prone to human error and increasingly ineffective against high-quality forgeries. These limitations underscore the urgent need for scalable, technology-driven solutions capable of proactive and automated counterfeit detection. Artificial Intelligence (AI), encompassing machine learning (ML), computer vision, and natural language processing (NLP), has emerged as a transformative tool in this domain. AI can rapidly analyse visual, textual, and spectral data to distinguish authentic from falsified medicines, identify anomalous packaging patterns, and detect suspicious product listings on e-commerce platforms [4]. For example, convolutional neural networks (CNNs) can capture micro-level inconsistencies in logo design or texture, while NLP models can monitor online sales channels for linguistic cues of fraud. Beyond standalone AI models, the integration of blockchain and distributed ledger technologies introduces a new paradigm for traceability and authentication. These systems enable immutable recording of manufacturing and distribution data, ensuring that every product’s digital identity can be verified across the supply chain [5]. When combined with AI-driven verification at the packaging or product level, blockchain provides end-to-end assurance from factory to patient. Scholar J Computational Science 104 Despite these technological advances, widespread adoption in public health remains limited by several challenges. Key barriers include data scarcity, particularly the lack of publicly available counterfeit datasets; algorithmic bias due to uneven data representation; and ethical considerations surrounding privacy, consent, and the misuse of surveillance tools. Moreover, implementation in LMICs faces logistical obstacles such as limited computational infrastructure, inconsistent internet connectivity, and varying regulatory readiness. This review aims to consolidate recent progress in AI-assisted counterfeit medication detection, focusing on applications relevant to public health and supply chain integrity. It critically examines four major technological domains: (i) AI and machine learning based detection approaches, (ii) synthetic and augmented data for model training, (iii) blockchain-enabled authentication frameworks, and (iv) mobile and edge-based deployments for decentralized verification. By synthesizing findings across these domains, the review highlights not only technical achievements but also the socio-ethical, infrastructural, and policy considerations necessary for responsible and equitable deployment. Ultimately, the review situates AI not merely as a technical innovation but as a public health enabler a tool capable of enhancing surveillance, strengthening regulatory enforcement, and empowering health systems to prevent the circulation of harmful counterfeit medicines. Review Methodology This review synthesizes peer-reviewed and preprint literature from 2018–2025, emphasizing AI, blockchain, and mobile-based approaches for counterfeit medicine detection. The focus is on studies integrating ML or computer vision models for authenticity verification, synthetic or augmented data for model generalization, and blockchain for traceability. Grey literature, including institutional reports and project briefs, was also reviewed to capture deployment insights. Studies were grouped thematically into four domains: i. AI/ML-based visual and spectral analysis, ii. synthetic and augmented data generation, iii. blockchain and distributed verification systems, and iv. mobile and edge-based implementations. The goal was not only to summarize individual contributions but also to critically analyze their methodological rigor, data sources, and implications for public health surveillance. AI and Machine Learning-Based Detection Approaches AI-driven detection systems primarily rely on machine learning models capable of distinguishing genuine from counterfeit products using either visual or chemical data. Among early examples, Alsallal et al. [4] developed an X-ray fluorescence (XRF) and ML-based approach to classify Tenormin tablets from multiple Scholar J Computational Science 105 manufacturers. Their system achieved 93% accuracy using a radial basis function support vector machine (SVM), demonstrating that spectral features can reveal subtle compositional differences between authentic and counterfeit drugs. However, while accurate in controlled conditions, such methods require laboratorygrade equipment and remain limited in scalability. More recent AI systems employ deep learning, especially convolutional neural networks (CNNs), for imagebased packaging verification. CNNs can learn complex spatial and texture features such as logo distortions, misalignment, or print inconsistencies—hallmarks of counterfeit packaging. Projects like MediVerify [5] integrate CNN classifiers with blockchain verification to authenticate pharmaceutical products using packaging and pill images. Although promising, such prototypes often rely on small or simulated datasets and lack benchmarking against large-scale, real-world samples. A related development is the verification-based approach proposed by Pandey et al. [6], which mirrors facial recognition systems. Rather than classifying images into “authentic” or “fake” categories, their model compares query images with reference authentic packaging via feature embeddings. This method improves interpretability and is resilient against unseen counterfeits. Nonetheless, its success depends heavily on the quality of reference databases, which are difficult to maintain across diverse global brands. Overall, AI-based models demonstrate high accuracy in laboratory evaluations but face challenges in field applicability. The main barriers include limited counterfeit datasets, lack of standardization across imaging conditions, and ethical concerns regarding patient or manufacturer data privacy. Synthetic and Augmented Data for Model Training Data scarcity is a central obstacle in counterfeit medicine detection. Authentic counterfeit samples are rare, and regulatory restrictions limit data sharing. Synthetic data generation has therefore emerged as a viable alternative to train and evaluate AI models without exposing proprietary or sensitive information. A recent policy framework from the Duke-Margolis Center for Health Policy [7] legitimizes the use of synthetic biomedical data, providing a structured pipeline for dataset generation, model training, and validation. The framework emphasizes maintaining statistical fidelity to real-world data while preserving privacy through controlled sampling and generative modeling. Motwani et al. [8] proposed a modular deep learning pipeline that leverages manually generated synthetic packaging images by modifying authentic samples (e.g., altering logos, fonts, or text). Their model combined object detection and optical character recognition (OCR) to extract visual and textual cues, comparing them against reference patterns using anomaly detection. The pipeline effectively demonstrated that even small, synthetically expanded datasets can significantly enhance generalization performance. Synthetic data has also been applied to spectral analysis. Wei [9] demonstrated the generation of simulated near-infrared (NIR) spectra by combining known chemical compositions, enabling the training of multiple ML models through ensemble learning. The synthetic spectral data allowed accurate counterfeit detection (up to 93%) without relying on real counterfeit samples. Scholar J Computational Science 106 While synthetic data enhances scalability and reproducibility, it introduces risks of overfitting to generated artifacts or lacking real-world variability. Consequently, hybrid approaches combining real, augmented, and synthetic samples currently offer the most balanced solution for counterfeit detection. Blockchain and Distributed Ledger Authentication Blockchain technologies provide immutable, transparent ledgers that can trace the provenance of pharmaceutical products throughout the supply chain. By recording manufacturing details, batch identifiers, and verification events, blockchain-based systems aim to make it difficult for falsified products to infiltrate legitimate distribution channels. Kumar et al. [10] proposed a blockchain-integrated traceability framework where each product unit carries a unique digital identity linked to its manufacturing and logistics history. End users can verify authenticity using QR codes or digital identifiers stored on the ledger. However, while conceptually strong, such systems often remain theoretical due to scalability and interoperability challenges, as well as the need for universal participation across manufacturers, distributors, and regulators. When integrated with AI, blockchain can provide a complementary verification layer. For instance, visual or spectral AI models can detect anomalies, while blockchain verification ensures the authenticity of supply chain metadata. Together, these dual mechanisms enhance trust, transparency, and security across the pharmaceutical lifecycle. Nevertheless, energy efficiency, latency, and regulatory harmonization remain barriers to global adoption. Mobile and Edge-Based Deployment in Public Health Contexts Mobile and edge computing platforms are reshaping field-based counterfeit detection, enabling real-time verification in low-resource or decentralized environments. Projects like MicroGuard [11] have developed smartphone-based AI tools capable of analyzing packaging and labeling features directly on-device, removing dependence on cloud infrastructure. This approach supports offline verification in areas with poor connectivity, making it particularly relevant for public health inspectors in LMICs. Similarly, commercial systems such as AlpVision’s Cryptoglyph and Fingerprint technologies [12] use invisible microdot patterns and surface texture signatures captured through standard smartphone cameras. These signatures serve as digital fingerprints, enabling authentication without specialized equipment. Such methods have strong potential for integration into public health supply chains, where fast, accessible verification can prevent counterfeit circulation at the community level. Mobile and edge-based solutions also align with global public health priorities for equity and accessibility. By lowering technical barriers, they allow local pharmacists, healthcare workers, and even patients to participate in decentralized surveillance. However, challenges include device heterogeneity, lighting sensitivity, and limited computational power for high-complexity models. Future work may involve federated Scholar J Computational Science 107 learning approaches that allow AI models to learn collaboratively from distributed devices while preserving data privacy. Discussion Artificial intelligence and digital technologies are reshaping how pharmaceutical quality assurance is conceptualized and implemented in global health. Across reviewed studies, a clear pattern emerges: while technical innovation in counterfeit detection has accelerated rapidly, translation into sustainable, scalable public health impact remains limited. The discussion that follows integrates insights from AI/ML-based detection, synthetic data methodologies, blockchain traceability frameworks, and mobile or edge implementations to examine how these technologies collectively address — and sometimes perpetuate — existing gaps in global medicine security. Convergence of AI with Traditional Analytical Frameworks Early work in counterfeit detection relied heavily on laboratory-based methods such as chromatography, spectroscopy, and X-ray fluorescence (XRF). These techniques, though accurate, required specialized expertise and infrastructure that limited their accessibility outside controlled environments [4]. The transition toward AI-enhanced analysis represents a paradigm shift from descriptive to predictive quality control. By automating pattern recognition and anomaly detection, machine learning models now enable nearinstantaneous evaluation of medicine authenticity using image or spectral inputs. Alsallal et al. [4] demonstrated this shift through the integration of XRF spectroscopy with machine learning classifiers, achieving over 90% accuracy in distinguishing counterfeit Tenormin tablets. However, such systems still depend on physical laboratory setups and cannot be scaled to remote or resource-limited regions. In contrast, computer vision models like those in MediVerify [5] and Pandey et al. [6] use smartphone-acquired images to verify authenticity, drastically reducing operational barriers. The transition from chemical to computational detection marks a major step toward democratizing pharmaceutical verification. Yet, these models remain constrained by the quality and representativeness of available training data. A further challenge lies in the interpretability of AI models. Public health agencies require explainable decision pathways to justify enforcement actions. Black-box neural networks, while accurate, often provide limited interpretability [3]. Bridging this gap demands a hybrid framework combining interpretable ML with traceable data sources, allowing regulators to verify both the decision logic and data provenance behind authenticity assessments. Data Limitations and the Role of Synthetic Data A central challenge across AI-based counterfeit detection studies is data scarcity. Regulatory restrictions, intellectual property protections, and the rarity of confirmed counterfeit samples all constrain the availability Scholar J Computational Science 108 of training datasets [7]. Without sufficient diversity, models risk overfitting to narrow feature distributions, failing when applied across new product batches, packaging designs, or geographic contexts. Synthetic data generation has emerged as an important countermeasure. Motwani et al. [8] manually created counterfeit packaging by altering authentic product logos and text, producing diverse image datasets that simulated real-world counterfeiting tactics. The resulting deep learning pipeline achieved robust generalization despite limited authentic samples. Similarly, Wei [9] generated synthetic near-infrared spectra based on known chemical profiles, demonstrating that models trained on simulated mixtures could detect falsified pharmaceuticals with accuracy comparable to models trained on real-world data. Policy frameworks such as the Duke-Margolis Center’s 2025 report [7] have begun formalizing synthetic data generation standards, providing legitimacy and ethical structure for their biomedical application. In the context of counterfeit detection, synthetic data not only expands model diversity but also supports privacy preservation by avoiding direct exposure of proprietary or patient-linked datasets. However, synthetic datasets are not immune to bias: generative models may replicate or exaggerate patterns from their training data, potentially reinforcing false associations. Future research should therefore emphasize hybrid data synthesis, combining real-world, augmented, and synthetic samples, validated through statistical fidelity checks and cross-domain testing. Blockchain Integration and Traceability Challenges Blockchain technology offers a complementary layer to AI-based detection by addressing the traceability problem in pharmaceutical supply chains. Whereas AI focuses on verifying the physical or visual authenticity of products, blockchain provides a digital trail of provenance [10]. In principle, every transaction—from manufacturing to dispensing—can be recorded on a distributed ledger, ensuring transparency and immutability. However, as Kumar et al. [10] observe, the theoretical appeal of blockchain often outpaces its operational maturity. Many pilot systems demonstrate proof-of-concept feasibility but lack evidence of scalability, interoperability, and user compliance. Integrating blockchain into national health systems requires alignment across regulatory frameworks, technical standards, and governance protocols. Additionally, energy consumption and data storage costs associated with large-scale distributed ledgers remain non-trivial, particularly in LMICs where computational infrastructure is constrained. From a public health perspective, blockchain’s greatest value may lie not in replacing regulatory oversight but in augmenting it—serving as a verifiable digital backbone upon which AI-driven inspection results, manufacturing certificates, and shipping records can coexist. The convergence of AI and blockchain could yield an end-to-end authentication ecosystem where image-based detection verifies physical characteristics while blockchain secures the data lineage, collectively deterring falsification at both the product and information levels. Scholar J Computational Science 109 Mobile and Edge-Based Innovations for Field Deployments While AI and blockchain provide the backbone of automated detection, mobile and edge systems operationalize these technologies in the field. Tools like the European Commission’s MicroGuard project [11] and AlpVision’s Cryptoglyph and Fingerprint technologies [12] exemplify how lightweight, smartphonecompatible solutions can bring pharmaceutical verification directly to pharmacists, customs officers, and even consumers. By leveraging on-device AI inference, these systems circumvent the need for cloud connectivity, enabling offline authentication—a vital feature for rural and low-connectivity regions. The move toward edge AI also supports privacy preservation, as sensitive image or product data never leaves the user’s device. However, practical challenges persist: lighting conditions, camera resolution variability, and hardware fragmentation can all impact detection accuracy. Continuous calibration and federated model updates, where devices collaboratively improve a shared AI model without exchanging raw data, represent promising solutions to these constraints. From a public health standpoint, the decentralization of detection empowers local stakeholders to become active participants in pharmaceutical surveillance networks. This shift aligns with global goals for community-based health monitoring and equitable access to digital tools. It transforms counterfeit detection from a centralized regulatory function into a distributed public health responsibility. Ethical, Regulatory, and Governance Considerations Despite their promise, AI-enabled counterfeit detection systems introduce new layers of ethical and governance complexity. Algorithmic opacity, data bias, and privacy risks are major concerns for health regulators [3], [7]. Moreover, in the absence of standardized performance metrics, different AI systems may yield inconsistent results, complicating legal adjudication and enforcement. There is also the issue of technological inequity: while high-income countries can deploy AI and blockchain infrastructures rapidly, LMICs often lack the digital backbone to support such systems. Without deliberate policy alignment and capacity-building efforts, the digital divide may inadvertently widen existing inequities in pharmaceutical surveillance. To mitigate these risks, international cooperation is essential. The WHO, International Medical Products Anti-Counterfeiting Taskforce (IMPACT), and regional regulatory alliances could play pivotal roles in setting interoperability standards, validating AI models, and providing certification pathways. Furthermore, ensuring transparency through open datasets, explainable AI algorithms, and community engagement will be crucial for maintaining public trust. Ultimately, the ethical deployment of AI in counterfeit medicine detection requires balancing innovation with accountability—aligning technological capacity with human rights, data protection, and equitable access principles. Scholar J Computational Science 110 Implications for Public Health Practice AI-enabled counterfeit detection extends far beyond pharmaceutical logistics—it represents a new public health intelligence frontier. Real-time analysis of packaging or distribution anomalies could serve as an early warning system for outbreaks of falsified medicines, allowing authorities to intervene before widespread harm occurs. By integrating AI surveillance data with national pharmacovigilance platforms, public health agencies could enhance risk prediction, improve medicine tracking, and strengthen consumer safety. However, to achieve this, digital systems must be embedded within existing health infrastructure rather than operate as parallel initiatives. Investments in workforce training, data governance, and cross-border interoperability will determine whether these technologies become sustainable tools for global medicine safety or remain isolated technical experiments. In summary, while AI, synthetic data, blockchain, and mobile technologies have shown tremendous promise in combating counterfeit drugs, their full potential will only be realized through coordinated, ethical, and inclusive implementation strategies. The next decade should focus not only on algorithmic innovation but also on institutional adaptation—ensuring that the future of pharmaceutical verification aligns with the overarching mission of public health equity and patient safety. Conclusion The growing threat of counterfeit medicines continues to undermine global public health, demanding innovative and scalable detection strategies. Artificial intelligence, synthetic data generation, blockchain, and mobile-based verification systems collectively represent a transformative shift from manual inspection toward intelligent, data-driven pharmaceutical surveillance. While AI improves accuracy and speed, blockchain ensures traceability, and mobile platforms expand accessibility in low-resource settings. However, realizing their full potential requires addressing challenges of data quality, model transparency, and regulatory integration. The future of counterfeit medicine detection lies in hybrid frameworks that unite these technologies under robust ethical, governance, and interoperability standards. By embedding AIdriven verification into public health infrastructures, health systems can move toward a more resilient, transparent, and equitable global supply chain. References [1] World Health Organization, “Substandard and Falsified Medical Products: Global Surveillance Report,” WHO, 2025. [2] S. Koh, Y. Tan, T. Lim, and C. Wong, “Advances in pharmaceutical serialization and traceability,” Int. J. Pharm. Sci., 2022. [3] S. Russell and P. Norvig, Artificial Intelligence: A Modern Approach, 4th ed., Pearson, 2021. [4] M. Alsallal, et al., “Counterfeit Drug Detection Using X-Ray Fluorescence and Machine Learning,” J.