Adversarial Attacks for Tabular Data: Application to Fraud Detection and Imbalanced Data

Guaranteeing the security of transactional systems is a crucial priority of institutions that process transactions, in order to protect their against cyberattacks and fraudulent attempts. Adversarial attacks novel techniques that, other than being proven to be effective to fool classification models, can also be applied to tabular data. Adversarial aim at producing adversarial examples, in other words, slightly inputs that induce the Artificial Intelligence (AI) system to return outputs that are advantageous for the attacker. In this paper we a novel approach to modify and adapt state-of-the-art algorithms to tabular data, in the context of fraud detection. Experimental show that the proposed modifications lead to a perfect attack success, obtaining adversarial examples that are also less perceptible when by humans. Moreover, when applied to a real-world production system, proposed techniques shows the possibility of posing a serious threat to the of advanced AI-based fraud detection procedures.