An Enhanced Learning Ensemble in Detection of Potential Threats via Anomalous Behaviour with Credit-Card Transactions

Abstract

The Internet as an effective model to advance resource sharing has consequently, led to the
greater proliferation of adversaries, with unauthorized access to network resources.
Adversaries achieve fraud activities via carefully crafted attacks of large magnitude
targeted at personal gains and rewards. With a cost of over $1.3Trillion lost globally to
financial crimes and the constant rise in fraudulent activities vis the use of credit-cards,
financial institutions and stakeholders must explore and exploit improved measures to
actively secure client data and funds. Financial services must harness the creative mode via
machine learning schemes to help effectively manage such threats. Our study thus, proposes
a cybersecurity machine learning XGBoost ensemble to detect fraud activities. This scheme
aim to equip a system with altruistic knowledge to help detect credit card fraud
transactions. Results show ensemble effectively differentiates fraudulent from genuine card
transactions with a model accuracy of 99.1%.