At Facephi, we see innovation as a collective effort: we move faster when we share what we know and learn from what is happening across the industry and the academic community. That is why we organized and evaluated the international document fraud detection challenge at IJCB 2026 together with Hochschule Darmstadt and Fraunhofer IGD, two of Europe’s leading institutions in biometric research.
The third edition of this competition (Competition on Document Forgery Detection on ID-Cards and Passports), held as part of IJCB 2026 (International Joint Conference on Biometrics), brought together more than 100 models from 63 teams representing universities and companies from around the world. Our R&D department, led by Javier Barrachina, evaluated each of these models on our platform—a volume that reflects both the global reach of the initiative and our technical capacity to process and validate algorithms at scale. And, as with any collaborative research effort, this year’s competition introduced passport evaluation for the first time, broadening the playing field for all participants.
The benchmark competition in presentation attack detection
Far from being just another competition, this challenge has established itself as the benchmark event for evaluating PAD systems on identity documents. Today, it is considered the state of the art in the field: no other initiative measures the performance of these algorithms in a neutral and controlled environment, putting academic and industry teams in the digital onboarding space on an equal footing.
We processed and analyzed the results together with Hochschule Darmstadt and Fraunhofer IGD. The former is home to the da/sec (Biometrics and Internet Security) group, one of the continent’s most prestigious academic centers for biometrics and security, led by Christoph Busch, a key figure in the international standardization of these technologies. The latter is part of Fraunhofer, Europe’s largest organization for applied research, and is a leading center in biometric security, with researchers such as Naser Damer. Working side by side with these institutions places us at the forefront of global research in digital identity.
The challenge was structured into two tracks. The first track was based on a unique synthetic image dataset, identical for all participants, with the aim of identifying the best algorithm for predicting presentation attacks under controlled conditions. The second track, which was open-ended, allowed each team to train using state-of-the-art data or, in the case of companies, their own proprietary data, and test their models against heterogeneous and more challenging attack scenarios.
The evaluation follows the international framework for measuring PAD systems (ISO/IEC 30107-3), which distinguishes between two types of errors: the proportion of attacks that the system fails to detect (APCER) and the proportion of genuine documents that it incorrectly rejects (BPCER). Based on these metrics, the competition ranks participants using an aggregate ranking—AV_Rank—which combines their performance across different decision thresholds, meaning that a lower value reflects better and more consistent performance.
Under these criteria, the winning team ranked first in both tracks, achieving an AV_Rank of 27.82% in the first and 68.71% in the second. The gap between these two figures is, in itself, one of the key takeaways from this edition: it illustrates just how much more challenging detection becomes when systems leave the laboratory and face diverse attacks and documents from different countries.
A new feature that brings the challenge closer to reality: passports
The main new feature of this edition has been the inclusion of passports in the set of documents evaluated. The two previous editions, starting with the first competition held at IJCB 2024, had focused exclusively on generic identity cards. Expanding the scope to passports brings the competition closer to real-world identity verification scenarios and significantly raises the technical demands placed on the algorithms.
Steady Improvement, but an Ongoing Challenge
Since its first edition, the challenge results have improved incrementally year after year, with error rates decreasing across both tracks. However, the competition also highlights a broader point: PAD systems are still not fully robust, and their performance continues to depend on the number of bona fide samples available, the type of attacks, and the variety of identity documents.
As of today, there is no universal system capable of predicting identity documents from any country with a low error rate. This is where the focus of research and development will lie in the years ahead.
«Organizing and evaluating a competition like this allows us to benchmark the best international algorithms in a neutral and controlled environment, using a common standard for everyone. The results confirm that document fraud detection is advancing every year, but also that it remains one of the major open challenges in digital identity. Only through collaboration between industry and academia will we be able to build truly robust PAD systems.»
Our participation in this initiative reaffirms our commitment to innovation and applied research as key drivers for strengthening security in identity verification processes on a global scale. In a context where document fraud is becoming increasingly sophisticated at the same pace as the technology designed to combat it, working closely with leading research institutions is at the core of our strategy.
The full results are available at: The Third Competition on Document Forgery Detection on ID-Cards and Passports.