Image Classification for Malware Detection using Extremely Randomized Trees

With the development of machine learning and computer vision, we combined the image classification and malware detection in this work. Based on the technology of image classification, we achieved the detection of malware, using machine learning. In our work, we visualized the malware as a grayscale image and extracted texture features by Gabor filter. Depend on machine learning, we used extremely randomized trees as the classification and 10-fold cross validation to value it. Compared with GBDT, KNN and RF, experimental results are quite impressive with the accuracy rate of 96.19% and the recall rate of 97.51% on a malware database of 15,781 samples.