Malware Detection in Image Data: A Deep Learning Approach

Malware is constantly changing in today’s digital environment, posing a rising danger to global cybersecurity. With the sheer amount and complexity of new malware, traditional detection techniques that frequently rely on signature-based approaches are finding it more and more difficult to keep up. To solve this problem, this study uses deep learning algorithms to detect malware hidden within images. The demand for quicker and more precise detection techniques is what spurred this research. The first step in the procedure is to put together a dataset that contains pictures of malware and innocuous content. We investigated image augmentation methods to increase the variety of the dataset. Next, using the processed dataset, a convolutional neural network (CNN) was created and trained. To guarantee a thorough assessment, pre-trained models such as VGGNet and ResNet were employed for comparison. 3813 images comprise the dataset used in this study; 2288 of those are training images, and the rest are divided into two halves as testing and validation images. With a detection accuracy of 97.38%, the trained model proved its capacity to identify images as either malware or non-malware. By offering a deep learning- based method for malware identification in images, this work advances the field of cybersecurity. It may make it easier to develop detection algorithms that are quicker and more accurate.