Development of Malware Classification Technique Based on Machine Learning Neural Network and High Order Markov Chain

Malware or Malicious Software is a computer program or software specially created and designed to interfere with, damage or gain unauthorized access to a computer system, which is part of a cybercrime activity. This is one of the main threats on the internet today, so a method needed to detect malware. Malware detection techniques are categorized into static methods and dynamic methods. Both of these techniques still have weaknesses, so machine learning is used to overcome the weaknesses of the two techniques. Convolutional Neural Network (CNN) is a type of neural network that is usually used to detect and classify malware in image data. However, in some classification tests with certain datasets, CNN still cannot work optimally. Therefore, a model development needed to improve the performance of CNN in doing classification. This thesis proposes a malware classification method using CNN using feature extraction from the high-order Markov Chain transition probability model as input for knowledge transfer. The purpose of developing this algorithm is to provide better classification results compared to conventional CNN methods. Various scenarios were carried out to get the best performance comparison results. The best performance was obtained when the model was run with the Markov transition probability order 1 dataset with 94% accuracy, 95% precision, 95% recall, and 95% f1-score.