Microsoft Malware Prediction Using LightGBM Model

The development of Windows malware detection technology and the evolution of malware are complementary and inseparable. Traditional malware detection methods mainly include signature-based detection, heuristic-based detection, and dynamic behaviour-based detection. Thanks to the development of computer hardware and artificial intelligence technology, machine learning and deep learning have achieved many remarkable research results in many fields, and researchers in network security are also paying more and more attention to machine learning. In our paper, we do feature engineering and use LightGBM as our model to process data. To evaluate our method’s performance, we use Auc-Roc as the metric. The higher the Auc-Roc Score and accuracy are, the better performance the model will gain. LightGBM model owns the highest Auc-Roc score 0.684, which is 0.046 and 0.007 higher than Catboost and Xgboost respectively.