Malware detection with machine learning methods

With the development of technology, the use of computer systems, which have become an integral part of our lives, is gradually increased. This increase has been caused many cyber security problems. Malicious software entering users' computer systems as a result of cyber security vulnerabilities are caused many damages. In order to prevent these damages, malware detection systems have been developed. It is vital for information systems that malware can to detected before it causes any damage. In this study, it has been estimated whether a malware exists on the system by using machine learning methods and telemetry information of the computer system. The data set used in the study was created by Microsoft by collecting telemetry information of computers. The first one million data rows of the data set were tested using 10 cross validation techniques with Naive Bayes, Decision Tree, Random Forest, Adaboost, LightGBM classification algorithms along with Knowledge Gain, Chi-Square feature selection methods. In addition, the data set obtained after data preprocessing steps by Lin (2019), who used the Microsoft Malware Prediction dataset in his study, was tested with Naive Bayes, Decision Tree, Random Forest, Adaboost, LightGBM classification algorithms after data preprocessing steps.