Malaria is a disease which, despite being present for over a century, still claims a significant number of lives every year. The advancement of artificial intelligence have opened the door to developing innovative methods in malaria treatment. Introducing machine learning approaches to this field can be beneficial in the disease prevention, detection, and therapy. In this work, convolutional neural networks for malaria detection are developed, based on the classification of thin blood smear images of the potentially infected cells. Input data was preprocessed using the image segmentation, file organization, image size standardization, color channel adjustment, and data splitting. Further, the proposed methodology included image conversion, network architecture defining, parameter tuning and network training. Various architectures of convolutional neural networks were developed and evaluated. In addition, multiple values of different network layer parameters were assessed. This study was implemented in Clojure programming language. Proposed network architecture includes two convolutional and pooling layers followed by activation functions, batch normalization and two linear layers. This convolutional neural network provided the best results and achieved an 82.7% accuracy. Furthermore, this paper proposes another network model with lightweight configuration and a slight accuracy decrease.