Malware Detection Using Semantic Features and Improved Chi-square

Malware Detection Using Semantic Features and Improved Chi-square

As advances in information technology (IT) affect all areas in the world, cyber-attacks also continue to increase. Malware has been used for cyber attacks, and the number of new malware and variants tends to explode in these years, depending on its trendy types. In this study, we introduce semantic feature generation and new feature selection methods for improving the accuracy of malware detection based on API sequences to detect these new malware and variants. Therefore, one of the existing feature selection methods is chosen because it shows the best performance, and then it is improved to be suitable for malware detection. In addition, the improved feature selection method is verified by using the Reuter dataset. Finally, the actual API sequences are extracted from the given malware and benign, and the proposed feature generation and selection methods are used to generate a feature vector. The performance is verified through classification.