Effects of Malware Detection Parameters on Classical vs Deep ML Techniques

Malware is one of the most significant cybersecurity threats during the era of digital transformation. An essential tool being used for malware detection is machine learning. A feature that researchers use extensively for machine learning-based malware detection is API calls. Our research attempts to determine how the number of API calls and the number of instances in the dataset affect the accuracy of the machine-learning models that are being used for malware detection. We test deep and non-deep learning models to answer these questions. Our research shows that those parameters significantly affect the deep learning technique with up to 20% reduction in the detection accuracy in detecting fake antivirus malware, whereas non-deep learning shows close to no impact with minor fluctuation in the detection accuracy.