Comparative Study of Prognosis of Malware with PE Headers Based Machine Leaning Techniques

As the threat of ransomware continues to grow, a conflict has broken out amongst those who are working to find and implement solutions. Yet, these systems are dynamic and continuously change owing to their reactive nature. Systems for early detection and protection have been developed and are now frequently employed. This is largely because malicious code may frequently be engineered to behave in ways that will mislead detection tools. Customers' devices must be protected against the annual influx of new malware. Users need to go beyond signature-based malware detection systems, which have issues recognizing zero-day ransomware, to secure their data from attacks caused by potentially harmful unknown ransomware. By analyzing the PE headers of a given piece of software, it is possible to extract relevant features that can be used to classify it as malicious or benign. In this paper, we evaluate the performance of various machine learning (ML) algorithms using a dataset of PE headers and compare the results to determine the most effective algorithm. Basically, we address a method based on ML for ransomware detection and make a comparative study. The experimental and testing results validate the efficacy of our method in detecting ransomware and differentiating it between harmful and benign. The results suggest that the usage of ML for malware prognosis can be useful as it can provide greater accuracy. The chi-squared test for feature selection proved to be improving accuracy in some cases as compared to ML methods without it. The findings of this study provide insights into the potential of using ML for malware prognosis and may inform the development of more effective cybersecurity solutions.