Image-Based Malware Detection Using α-Cuts and Binary Visualisation

Image conversion of malicious binaries, or binary visualisation, is a relevant approach in the security community. Recently, it has exceeded role single-file malware analysis tool and become part Intrusion Detection Systems (IDSs) thanks to adoption Convolutional Neural Networks (CNNs). However, there been little effort toward image segmentation for converted images. In this study, we propose novel method that serves dual purpose: (a) enhances colour pattern segmentation, (b) achieves sparse representation According this, considered R, G, B values each pixel as respective fuzzy sets. We then performed α-cuts defuzzification across all pixels image, which them matrices 0s 1s. Our was tested on variety dataset sizes evaluated according detection rates hyperparameterised ResNet50 models. findings demonstrated larger datasets, representations intelligently coloured images can exceed model performance unprocessed ones, with 93.60% accuracy, 94.48% precision, 92.60% recall, 93.53% f-score. This first time were used processing our results, believe they provide an important contribution challenging datasets. Overall, shows integrated component image-based IDS operations other demanding real-time practices.