Malware Variants Detection Based on Feature Fusion.

Being able to detect malware variants is an important problem due to the rapid development and the security threats of new malware variations. Machine learning methods are currently one of the most popular malware variant detection methods, however, most of these methods only use single type of features (e.g. opcode) and shallow learning algorithms (e.g. SVM), which also makes these methods have demonstrated poor detection accuracy and low detection speeds. In this paper, we propose a method that combines multiple features of malware with deep learning methods to optimize the detection of malware variants. To implement the proposed method, we use Deep Convolutional Neural Network (DCNN) and Information Gain (IG) to extract effective features from the grayscale map and disassembly file mapped from the malware, respectively. Then we construct a fusion feature space by combining the different types of extracted features and use it to train a Multilayer Perceptron (MLP) to obtain results. The experimental results demonstrated that our method achieved good accuracy as compared with other common malware detection methods.