Improving Safety in Europe - Detecting Counterfeit Certificates from FFP2 Masks: A Text-Mining Approach

At the beginning of the Corona pandemic, there was a surge in demand for medical masks, especially FFP2 masks, which could not be adequately met by existing mask manufacturers. This led to new manufacturers of FFP2 masks entering the market. However, in order for these new suppliers to be allowed to offer these products in conformity with the law, the masks must be tested by an independent body (notified body). This testing usually takes several months and is associated with a number of bureaucratic hurdles and expenses. This led to the fact that untested masks were sold and are still being sold. In order to subsequently identify untested masks, we propose a method that automatically checks a specific document from the testing, namely the type examination certificate. To enable automated evaluation, this paper presents a text-mining approach for verification. It consists of three steps: i) requirements identification of certificates, ii) data extraction and analysis and iii) evaluation of the results using a decision tree classifier. Finally, the algorithm is validated using a case study of several counterfeit FFP2 type examination certificates. The result shows that the text-mining approach is capable of distinguishing fake from real certificates. The ability to analyze certificates in an automated way will completely change the way market surveillance authorities and accredited verification bodies check documents.