Document Fraud Detection at the Border: Preliminary Observations on Human and Machine Performance

How many false documents (forged and counterfeited) cross the border undetected? What is the real extent of the phenomenon of document fraud at the border? How good are border officers and their technical equipment at detecting fraud in the first line? How can we measure the impact of capacity-building measures (training and technology) in the field of document fraud detection? This paper proposes to approach the traditional problem of "known unknown" of risk analysis by taking the performance of detection capabilities - human and machine-supported - seriously. It argues that capability-based vulnerabilities need to be systematically assessed quantitatively and qualitatively in order to make sense of the risk and to devise, test, and measure the effectiveness of countermeasures. The paper reports the preliminary results of an exercise simulating the first line of document inspection at the border. European document experts and automated document inspection systems were challenged to recognize genuine and false documents under a very tight time constraint. Although the experiment suffered of many methodological weaknesses due to the limitations of the context in which it was conducted, a number of initial observations can be drawn on the importance of human skills and experience, the strengths and shortcomings of automated systems, and the need to further test and study how human and machine capabilities can be improved and combined in order to increase their detection effectiveness and thus strengthen border security.