Deep Learning Based Malware Classification Using Deep Residual Network
Document Type
Conference Paper
Publication Date
2019
Publication Title
Proceedings, MSVSCC 2019. 13th Annual Modeling, Simulation & Visualization Student Capstone Conference
Pages
126-131
Conference Name
MSVSCC 2019. 13th Annual Modeling, Simulation & Visualization Student Capstone Conference, April 18, 2019, Suffolk, VA
Abstract
The traditional malware detection approaches rely heavily on feature extraction procedure, in this paper we proposed a deep learning-based malware classification model by using a 18-layers deep residual network. Our model uses the raw bytecodes data of malware samples, converting the bytecodes to 3-channel RGB images and then applying the deep learning techniques to classify the malwares. Our experiment results show that the deep residual network model achieved an average accuracy of 86.54% by 5-fold cross validation. Comparing to the traditional methods for malware classification, our deep residual network model greatly simplify the malware detection and classification procedures, it achieved a very good classification accuracy as well. The dataset we used in this paper for training and testing is Malimg dataset, one of the biggest malware datasets released by vision research lab of UCSB.
Original Publication Citation
Lu, Y., Graham, J., & Li, J. (2019). Deep learning based malware classification using deep residual network. In Proceedings, MSVSCC 2019. 13th Annual Modeling, Simulation & Visualization Student Capstone Conference (pp 126-131). Old Dominion University. https://digitalcommons.odu.edu/cgi/viewcontent.cgi?article=1016&context=msvcapstone#page=133
Repository Citation
Lu, Yan; Graham, Jonathan; and Li, Jiang, "Deep Learning Based Malware Classification Using Deep Residual Network" (2019). Electrical & Computer Engineering Faculty Publications. 373.
https://digitalcommons.odu.edu/ece_fac_pubs/373
ORCID
0000-0003-0091-6986 (Li)
Comments
Link to proceedings as a whole: https://digitalcommons.odu.edu/msvcapstone/proceedings/2019/1/