As digital marketing becomes more and more crucial in today’s business strategies, the issue of click fraud. Which aims to deplete advertising budgets through activities. Has grown, leading to the need, for defenses. This research introduces a blend of Convolutional Neural Networks (CNN) Bidirectional Long Short Term Memory networks (BiLSTM) and Random Forest (RF) classifiers to effectively combat click fraud. This innovative model combines the power of learning for extracting features with RFs classification capabilities. By incorporating CNN and BiLSTM our method improves the identification of spatial features from data, which are then classified using an RF model. This ensemble approach is designed to address shortcomings in existing models that depend mainly on machine learning methods and often demand extensive feature manipulation. We showcase the efficiency of our model through testing on a dataset, achieving an impressive accuracy rate of 99.19% surpassing previous techniques. Additionally, we evaluate metrics such as precision, recall, and F1 score all highlighting the superiority of our approach. This research does not only push forward the tools for detecting click fraud but also offers a scalable framework for potential adaptation and implementation, in real world scenarios.