Detection of malware based on the classification of source code graphs

Abstract: This paper contains an analysis of existing approaches to de- tect the malicious software. An approach to classify the soft- ware in computer systems of information processing is of- fered. Proposed approach consists of three steps: construction of the flow graph of an application's source code, vectorization of the received graph and classification of the graph based on an artificial neural network of adaptive resonance theory. The graph is vectorized based on the bag-of-word model. Obtained results show an application of proposed approach in malware detection field.

Keywords: malware detection, control-flow graph, neural network, data classification