A Comparison Study to Detect Malware using Deep Learning and Machine learning Techniques
Biodoumoye George Bokolo, Razaq Jinad, Qingzhong Liu
Abstract
Malware creation has evolved from basic malware that is easy to detect to complicated malware that is obfuscated and quickly adaptive, raising the challenge of being easily detected. This study compares seven machine learning and deep learning techniques in detecting malware by using the extracted byte, opcode, and section codes. In this research, we aim to classify malware in nine different malware families correctly. First, the byte codes, the section codes, and the opcodes of the different malware applications are extracted and merged, and the classification is done by using the Random Forest, Decision Tree, Support Vector Machine, K-nearest neighbor, SGD, Logistic Regression, Näıve Bayes, and deep learning techniques. The result shows that the deep learning model outperforms other compared machine learning algorithms with an accuracy of 96%. Overall, the paper highlights the importance of using advanced machine learning and deep learning techniques for the detection of malware, particularly given the increasing complexity and adaptiveness of modern malware. The findings of the paper suggest that deep learning techniques may be particularly effective for detecting and correctly classifying malware.