Detection of malware by static analysis using machine learning and deep learning methods

Malware detection is an important aspect of computer system security. In the internet era, malware (such as viruses, trojans, ransomware, and bots) has posed significant security threats to internet users. The variety of malware is growing rapidly and exponentially. This situation brings with it the necessity of using new techniques for the correct identification and detection of malicious software. Deep Learning (DL) and Machine Learning (ML) models are powerful technologies that are transforming the effectiveness of cybersecurity research. The data used for the training of the model in the study consists of harmful and harmless software obtained from C-Prot company. Within the scope of static software analysis, 1000 harmless software and 1000 malicious software were examined without running the software, and the functions, libraries, digital signatures and other features used by the software were analyzed and the functional structure of the software was extracted. Machine Learning and Deep Learning methods were used and the obtained results were evaluated. Calculated accuracy, f1-score, precision, and recall metrics are compared to determine success of the methods. It has been observed that the Long Short Term Memory Networks algorithm, which is one of the deep learning methods, gives the most successful result with an accuracy rate of 99.75% among the methods used.