Novel Hybrid, Feature-Based Malware Detection Model Using Machine-Learning Techniques

Our research introduces an efficient malware detection model that uses machine learning to tell apart safe and harmful executable files. We aim to combat the increasing malware threat with a straightforward yet effective approach. Our method combines two types of analysis: static and dynamic. We want to compare the results from both and create a better, hybrid method for more accurate detection. Our dataset contains malware and safe executable samples, which we use to train and test our model. We pay close attention to extracting useful information from file and section headers of portable executable files. For classification, we try different machine learning classifiers like Support Vector Machine, AdaBoost, XGBoost, Random Forest, KNearest Neighbors (KNN), and Bernoulli Naive Bayes (BernoulliNB), k-means. The standout performer is the Random Forest classifier, achieving an impressive 98% accuracy rate. This research shows that machine learning is effective in spotting malware and emphasizes the importance of picking the right classifier for the job.