Malaria remains one of the most pressing global health challenges, particularly in endemic regions where a timely and accurate diagnosis is crucial. In this work, we propose a hybrid deep learning framework that combines advanced image pre-processing, a custom convolutional neural network (CNN) for feature extraction, an enhanced feature selection mechanism based on Levy-flight Grey Wolf Optimization (GWO), and a support vector machine (SVM) classifier. The proposed methodology is designed to mitigate the limitations of manual microscopy and conventional computer-aided diagnosis, achieving superior detection accuracy while reducing computational overhead. Experimental evaluation on multiple publicly available malaria datasets demonstrates an accuracy exceeding 97%, outperforming several baseline deep learning architectures. We discuss the strengths, challenges, and future potential of integrating domain-specific pre-processing with modern optimization techniques in medical image analysis.