On May 22nd, Work Package 5 of the TITAN Project hosted their first webinar wherein Lylia Abrouk, associate professor at The University of Burgundy presented the methods she uses to detect inconsistencies in data, focusing on possible instances of fraud. Lylia outlined three main approaches to fraud detection:
- Rule-based approach: This involves imposing logical constraints on the data to test its legitimacy and likelihood of being fraudulent.
- Machine learning: Algorithms are employed to predict the likelihood of a given instance being fraudulent.
- Ontology-based approach: This method incorporates an agreed-upon set of vocabulary to combine the strengths of the rule-based and machine learning approaches.
It is also possible to combine these approaches for enhanced fraud detection capabilities. Lylia emphasized the challenge of obtaining sample datasets containing known instances of fraud, which hampers the development of more efficient methods. Consequently, her presentation served as a plea for relevant data that can aid in the adaptation of existing fraud detection techniques.
Following the presentation, the 27 participants had the opportunity to ask Lylia questions. They expressed interest in applying these methods to agro-food supply chains. Lylia noted that although her examples were based on financial transactions, these approaches can be adapted to other data types within agro-food supply chains. The crucial factors for success are the quality and quantity of the data, as well as collaboration with domain experts familiar with various industries and products. This collaborative effort ensures that suitable solutions are developed for different supply chains and products, as the methods need to be tailored to each specific data type.
A recording of the webinar was created and can be made available upon request. For any further inquiries, please contact Pascal Neveu and Bernadette Durand at INRAe.