|
Original research AN EFFICIENT MALWARE DETECTION MODEL THROUGH REDUCED HYBRID FEATURES AND ENSEMBLE MACHINE LEARNINGPages 113-122
Abstract
Malware attacks are increasing in a higher rate due to extensive use of Internet & handheld devices and misuse of technological advancements. There are many malware detection techniques which use static or dynamic features for classification. This work focuses on creating static feature vector, dynamic feature vector, combining them to form a hybrid feature set and prepare it for better classification. Printable string information and API call sequences of malware and benign samples are used as two feature sets which are passed through feature selection algorithm for selecting best features. The reduced feature sets are combined to form a hybrid feature set. The hybrid feature set is passed through an ensemble model for classification. The ensemble model used in this article includes three supervised classifiers such as SVM, KNN and DT. This proposed model has performed well compared to the individual classifiers as well as individual feature sets. Besides better accuracy, this model has an execution time approximately equal to the individual classifiers which makes this model efficient for malware detection.
Keywords: Malware detection, static analysis, dynamic analysis, Malware API calls, Ensemble Learning, Feature Selection.
|
Om Prakash Samantray, 
PDF 