Detecting Malware with an Ensemble Method Based on Deep Neural Network

摘要: Malware detection plays a crucial role in computer security. Recent researches mainly use machine learning based methods heavily relying on domain knowledge for manually extracting malicious features. In this paper, we propose MalNet, novel malware method that learns features automatically from the raw data. Concretely, first generate grayscale image file, meanwhile its opcode sequences with decompilation tool IDA. Then MalNet uses CNN and LSTM networks to learn sequence, respectively, takes stacking ensemble classification. We perform experiments more than 40,000 samples including 20,650 benign files collected online software providers 21,736 malwares provided by Microsoft. The evaluation result shows achieves 99.88% validation accuracy detection. addition, also take family classification experiment 9 families compare other related works, which outperforms most of works 99.36% considerable speed-up detecting efficiency comparing two state-of-the-art results Microsoft dataset.