Fraud especially smartcard and telecommunication-based fraud always leave a grievous loss to its victims. The banking sector and the telecom companies has battled with this
plague for years, fighting it with both technological and other security measures to eliminate its occurrence, but there are still open- problem despite all the efforts. Most of the systems developed are usually reactive instead of proactive, i.e. they detect the fraud after they have already occurred instead of preventing it, and others detect the fraud but do not have the
mechanism to prevent it from occurring. Hence, the need for the development of an enhanced model that can detect Smartcard and telecom frauds in real-time and block the transaction while informing the relevant stakeholders (the account owner and the bank). In this work, we used the neural network to train a system using historical dataset of credit card fraudulent transactions and telecom fraud. This system could eliminate the inefficiencies of the existing systems and produced more efficient fraud detection and prevention using the Rule-Based approach to classify suspicious transactions and flag them if they contradict the rules. The result shows that fraud detection can now be made using well prepared datasets. Our model scored a performance accuracy mark of 94% as opposed to the existing system which had 65%. This work could be beneficial to telecom industries, to banks, to users of smartcards and POS and to every other person who carries out cashless transactions via the smartcards.