This study centres on dynamic malware detection, recognizing that malicious software evolves continuously and demands more adaptive security approaches. With new malware emerging almost every day and exploiting weaknesses across the Internet, traditional manual and heuristic-based analysis has become insufficient. To address this growing challenge, this research employs automated, behaviour-based detection supported by machine learning techniques. In this approach, malware samples are executed within a controlled environment, their behaviours are monitored, and detailed reports are generated. These reports are then transformed into sparse vector representations, which serve as input for various machine learning models. The classifiers applied in this study include kNN, DT, RF, AdaBoost, SGD, Extra Trees, and Gaussian NB. An evaluation of the experimental results shows that RF, SGD, Extra Trees, and Gaussian NB all reached 100% accuracy on the test set, along with perfect precision (1.00), recall (1.00), and f1-scores (1.00). These findings suggest that a proof-of-concept system combining autonomous behaviour analysis with machine learning can detect malware both effectively and efficiently.
Singh et al. (2026) studied this question.