Concept for Automated Computer-Aided Identification and Evaluation of Potentially Non-Compliant Food Products Traded via Electronic Commerce

Krewinkel, A., Sünkler, S., Lewandowski, D., Finck, N., Tolg, B., Kroh, L. W., … Fritsche, J. (2016). Concept for automated computer-aided identification and evaluation of potentially non-compliant food products traded via electronic commerce. Food Control, 61, 204–212.

Download: Preprint / Publisher Version


The online marketplace for food products is continually expanding and all types of food and beverages can now be purchased over the internet. It is primarily the responsibility of the food business operator to ensure compliance with food safety law. However, the competent authorities are tasked with controlling the e-food sector as part of their regulatory duties to protect consumer health and to prevent fraud, regardless of the sales channel being used. For this purpose, a new software prototype concept was developed that automatically identifies and evaluates potentially non-compliant e-food products. The prototype was developed using a modular architecture comprising a research tool, an image analysis tool, and a monitoring tool. User-defined thresholds are applied to assess the reliability of the retrieved data. Results that are not deemed reliable enough can be reworked using a computer-aided evaluation interface. The research tool utilizes both internet search engines and customized search algorithms. A multi-stage filtering process is performed to limit the sites according to defined criteria (e.g. food product merchants only). The data acquisition module stores all matching data from webpages for later analysis and preservation of evidence. In another module, automatic recognition of a site’s legal notice (impressum) is carried out for the respective vendor within whose online shop a potentially non-compliant food product is being offered. The image analysis tool performs logo recognition to enrich the text-based information of websites, thus providing additional visual information. The monitoring tool performs regular automated monitoring of e-food vendors, products and ingredients. The proof of principle of the prototype was achieved by conducting a web search for hazardous food products containing synephrine and caffeine. In total, 1242 product offerings on the internet for suspicious food products were identified among the 8683 search results. The software prototype has potential to enhance consumer protection and food safety with respect to e-foods.