Malware Detection System Based on Static-Dynamic Preprocessing Techniques Combined in an Ensemble Model

Malware distributors use persistence mechanisms in their malware to keep a PC infected for a long time. Techniques such as packing and obfuscation are often used so that the presence of malware cannot be determined by the simple executable code of the program. Static analysis of executable code, structures, and strings cannot determine obfuscated and packed malware. Therefore, dynamic analysis is required to determine the malicious behavior imposed on the system. In this paper, we used the CNN (Convolutional Neural Network) and LSTM (Long Short-Term Memory) model, which preprocessed static data and dynamic data and combined them into an ensemble model to determine malicious and benign and achieved an accuracy of 97.68%, F1-Score of 97.67%, and false positive rate of 1.57%.