Leader Talks

Beyond the QR Code: Can Physical-Item Authentication Strengthen

Imagine a pharmaceutical pack carrying a valid QR code and serial number. It passes verification, but what if the code has been copied onto a counterfeit pack? How do you authenticate the physical pack itself?

This is the gap Blue Cromos is addressing through AI-powered computer vision, using the unique physical characteristics of packaging materials to create digital fingerprints for individual packs. In this edition of Hello Pharma, Mete Sengul, Computer Vision Engineer and R&D Lead at Blue Cromos, discusses how physical-item authentication works and how it could strengthen existing pharmaceutical anti-counterfeiting systems.

1. Pharmaceutical packaging relies heavily on printed codes and serialization. How can the physical characteristics of packaging materials provide an additional layer of product authentication, and what determines the reliability of this approach across different materials and packaging formats?

Printed codes identify a pack, but anyone with a printer can copy them. Physical-item authentication adds a second layer by recording features of each individual pack that cannot practically be reproduced. These include the microstructure of the carton board or the tiny variations in how each label and code were printed. The reliability of this approach depends mainly on how much unique, stable structure the material has and how well it can be imaged. Uncoated board and paper labels are rich in texture, while glossy varnishes, foils and transparent films reflect light and require a tailored imaging setup.

2. Printing, coating, handling and transportation can affect packaging surfaces. How should manufacturers assess the impact of these variables on authentication accuracy, and what testing is necessary to establish reliable performance under real-world conditions?

We recommend evaluating each packaging design separately, using many different packs rather than many photos of just a few. Testing should mimic the real life of a pack, including abrasion and scuffing, humidity and temperature, simulated transport and aging. The same packs should be checked before and after these tests. Two numbers matter most, both measured at the threshold that will be used in production: how often a genuine pack is recognised and how often an unregistered or counterfeit pack is correctly rejected.

3. What practical challenges arise when integrating physical-item authentication into pharmaceutical packaging lines, particularly in terms of line speed, changeovers, equipment compatibility and validation? How do these considerations differ between brownfield and greenfield facilities?

The main challenges are capturing a detailed, sharp and well-lit image of every pack at full line speed, handling changeovers without re-engineering, and fitting the system alongside existing serialization and vision equipment.

Our approach uses a capture station positioned after the serialization station. As each pack passes through, it is photographed and its serial number is read from the same image that captures the pack’s unique physical characteristics and label. This links the serial number directly to the pack’s physical fingerprint, avoiding a separate matching step. Both are stored together as the reference for that individual pack, which is retrieved through the serial number during subsequent verification.

The capture station uses standard industrial cameras that can be adjusted to suit different line speeds and products. Validation follows the familiar IQ/OQ/PQ path. In brownfield sites, the main constraints are physical space and integration with existing serialization stations. Greenfield facilities can plan the capture point, lighting and data flows right from the design stage.

4. When counterfeiters copy serial numbers or reuse genuine pharmaceutical packaging, how can physical-item authentication help identify potential counterfeits or product substitution? How can it complement existing serialization and track-and-trace systems in these scenarios?

A copied serial number can verify correctly in a repository until a duplicate scan raises an alert, which may be too late. Physical-item authentication ties the code to the physical pack it was printed on, so a valid number appearing on the wrong pack can be flagged at the first check.

Reused genuine packaging is harder to detect because the pack itself is real. In this case, the value comes from also checking tamper-evident features, such as detecting whether a seal has been opened or replaced.

It complements serialization rather than replacing it. The serial number identifies which pack it should be, while the physical check confirms that it really is that pack and that it is intact.

5. What validation, data integrity, audit trail and documentation requirements should manufacturers address before relying on physical-item authentication during counterfeit investigations or quality assessments? How can they ensure the resulting evidence is traceable and defensible during regulatory inspections?

Before relying on it in investigations, manufacturers should validate the system for its intended use, just like any other computerized system in a regulated plant. This includes requirements such as the FDA’s 21 CFR Part 11 and EU GMP Annex 11.

In practice, every result should be stored with its source images, the time, device and operator details, and the exact model and settings used. Every change should also leave an audit trail. This allows an investigator to see exactly why a pack was flagged, making the evidence traceable and defensible during an inspection.

6. As manufacturers expand authentication across multiple SKUs and production sites, how should they evaluate scalability, implementation costs and long-term maintenance? Looking ahead, how could integration with MES and digital product passports strengthen product provenance, and what data governance and cybersecurity challenges must be addressed?

Scalability depends on how much work each new product needs. How reliably packs are recognised depends mainly on the packaging material and printing. Products that use the same packaging can therefore share test results instead of each requiring a full new evaluation.

Cost assessments should look beyond the software itself. Manufacturers need to consider capture hardware for each line, integration and validation, maintenance when packaging designs change, and storage for the reference data of every registered pack. This data grows with production volume and may need to be retained for years.

Looking ahead, connecting to the plant’s Manufacturing Execution System (MES) could start registration automatically for each batch and link it to the batch record.

The EU’s digital product passports won’t cover medicines, but serialization already plays a similar role in pharma. Physical-item authentication adds a physical link to this.

The main challenges are controlling access to the reference data and protecting it, since it is effectively the key to recognising genuine packs.

About the Guest

Mete Sengul is Computer Vision Engineer and R&D Lead at Blue Cromos, specialising in AI-driven computer vision technologies. His work focuses on developing AI-powered digital fingerprinting solutions that analyse the unique physical characteristics of products, enabling individual item authentication and helping address the growing challenge of counterfeiting.

Disclaimer

The views expressed in this editorial are those of the interviewee and are based on his professional experience in AI, computer vision and physical-item authentication. The explanations of Blue Cromos’ technology and its potential applications have been provided by the interviewee and have not been independently validated by Hello Pharma.

This editorial is intended for informational and educational purposes only. It does not constitute an endorsement or recommendation of any specific technology, product or solution by Hello Pharma. Readers are encouraged to independently assess the information and its applicability before making professional or operational decisions.