A multi-skilled team providing protection for the full range of payment methods used by our banking clients 

Crédit Agricole Payment Services is involved in fraud detection for all the Crédit Agricole Group’s banks. It intervenes in the areas of e-payments, transfers, authentication and, soon, checks. It aims to protect clients against the fraudulent use of their means of payment while limiting the number of false alarms generated which they may find irritating. A team of business line experts and two data scientists works full time to achieve this and is constantly developing innovative methods by drawing on fraud and artificial intelligence expertise. Payment fraud involving the use of bank cards and money transfers totalled €740 million in France in 2020(1), and AI is increasingly becoming a major development priority when it comes to protecting the means of payment used by our clients. 

AI has a growing role to play in our anti-fraud procedures 

Research into AI for the purpose of tackling fraud began in 2019 and was aimed at adapting the complexity of the algorithms being developed to the constraints of real-time fraud detection. The cumulative number of card payments and money transfers made in France in 2020 exceeded 18?billion transactions(2), so CAPS performs a risk analysis on several hundreds of transactions every second. AI algorithms must also be up to the major challenges of adaptability, robustness and performance. These algorithms (segmentation, decision tree and metaheuristics(3)) have considerably increased AI’s contribution to fraud detection efforts, from 2% to 12% in one year! 

Research to continually improve our anti-fraud procedures 

Fraud behaviours are becoming increasingly sophisticated and complex, so Crédit Agricole Payment Services aspires to be at the cutting edge of technology when protecting the Group’s banking clients. A current research partnership with  Ecole polytechnique (top French engineering school) is studying the automatic identification of new fraud behaviours. This research partnership builds models, factors in fraud behaviours and predicts future fraud behaviours using deep learning(4). 

(1)Source: Observatory for the Security of Payment Means, 2020 annual report. Payment fraud by bank card in France amounted to €473?million in 2020 versus €267?million for money transfers.? 
(2)Source: Observatory for the Security of Payment Means, 2020 annual report. Some 14 billion bank card transactions were made in France in 2020 versus over 4?billion money transfers.? 
(3)General algorithmic approximation methods that can rapidly provide a feasible albeit not necessary optimal solution to a complex optimisation problem.? 
(4)A machine learning method based on the principle of neural networks by stacking up a large number of “neuron” layers in order to establish complex rules. This method seeks to mimic the human brain.?