A New Learning Approach to Malware Classification Using Discriminative Feature Extraction

Authors:

S. Nandakishore, Dr. M. Arathi

Page No: 1283-1289

Abstract:

Starting from the presentation of the Web, malware has formed into perhaps of the most serious danger. A crucial step towards effective elimination is recognizing distinct malware types. Malware is transformed into a picture for the purpose of visualization and classification in malware visualisation, a subset of malware static analysis methods. Regardless of critical advancement, extricating fitting surface element portrayals for intense datasets stays troublesome. Global picture attributes that are sensitive to relative code positions are used in the methods that are currently in use. We present a smart learning strategy in this survey to make more discriminative and generous part descriptors. The proposed methodology uses existing close by descriptors, for instance, neighborhood equal models and thick scale-invariant component changes, gathering them into blocks and using one more bag of-visual-words model to convey generous features that are more versatile than overall components and more solid than adjacent features. Three malware datasets were utilized to test the proposed strategy. The aftereffects of the examinations show that the inferred descriptors have state of the art arrangement abilities.

Description:

malware Detection, GIST and SIFT image features, combined decision, machine learning, malware Analysis.

Volume & Issue

Volume-12,Issue-4

Keywords

.