Malware Detection Techniques and Research Based on Image Analysis

With the development of Internet, the speed of malware iteration is accelerating. To cope with the new scale and rapid variation, further optimize the model structure of malware detection, and improve the detection efficiency and accuracy, we establish a malware detection model based on image analysis. The steps of the model are as follows: (1) Image the malware into gray-scale image through B2M algorithm, and organize the malware gray-scale map dataset; (2) Establish the Keras-based CNN model and fill the model for training and testing; (3) Save the model parameters to establish malware Detection model. We introduce the image analysis technology and convolutional neural network correlation theory, as well as the construction steps of the detection model, and then evaluates and analyzes the detection results, compared with traditional image vector-based PCA methods and LDA methods, singular value decomposition methods for image feature extraction, and other algorithms. The model has efficient structure, which realizes lightweight with low time and spatial complexity, improves the efficiency of detection while maintaining high detection accuracy, and also has good detection capability for variant malware. Currently, it can better cope with the new characteristics of large-scale and rapid transformation of malware, which has a meaningful future for development.