Real-Time Ad Click Fraud Detection Using Attention Enhanced CNN Architecture

Abstract

Ad click fraud poses a significant challenge to online advertising platforms by generating illegitimate clicks that distort campaign performance and increase financial losses. To address the limitations of conventional machine learning and deep learning models in capturing complex contextual patterns, this work presents an attentionbased Convolutional Neural Network (CNN) framework for real-time ad click fraud detection. The proposed extension integrates an attention mechanism within the CNN architecture to emphasize the most informative features and sequential dependencies in clickstream data, thereby improving prediction reliability. Additionally, a webbased deployment using the Flask framework enables real-time fraud prediction through an interactive interface. Experimental evaluation demonstrates that the attention-enhanced CNN achieves superior accuracy and robustness compared to baseline models, highlighting its effectiveness for scalable and practical ad click fraud detection in dynamic online advertising environments.