The widespread use of the Internet has led to an exponential increase in cybersecurity attacks. One of the biggest threats is malicious software, also known as malware, which is increasing in number and sophistication. As existing techniques for malware detection are often unable to keep pace with evolving malware threats, the need for robust malware detection techniques has become paramount. In this paper, we present MalDualCNN, a novel malware detection technique utilizing binary file visualization and deep learning. The visualization of binary files serves as a valuable tool in analyzing, understanding, and depicting the behavior exhibited by malware. We introduce a novel technique for visualizing Windows executable files using locality-sensitive hashing. This technique converts each Windows executable file into two separate informative color images, whereby executable files with similar characteristics exhibit similar color images. The potency of convolutional neural networks, or CNNs, enables us to effectively learn similar patterns and distinctive features for the malware detection task. To serve this purpose, we develop a CNN model that employs two parallel convolutional blocks, each focusing on distinct aspects of an executable file. Our experimental results demonstrate that MalDualCNN can effectively detect malware with an accuracy of more than 97%, outperforming state-of-the-art techniques in terms of detection performance.