Malware Feature Extraction of the Alibaba Cloud Malware Dataset via Machine Learning Algorithms

This research explores a method for detecting and analyzing malware using machine learning algorithms, specifically variable-length N-grams and XGBoost, to extract significant behavioral features from API call sequences. By preprocessing the dataset and conducting statistical analysis to identify differences in API occurrences among malware families, the proposed method demonstrates high accuracy in classifying malware samples from the Alibaba Cloud dataset. The results show a 91.3% accuracy for multiclass classification and 99.0% for binary classification, highlighting the potential of variable-length N-grams and XGBoost in enhancing malware detection and analysis.