Fast and Efficient Malware Detection with Joint Static and Dynamic Features Through Transfer Learning

Page view(s)
30
Checked on Feb 17, 2025
Fast and Efficient Malware Detection with Joint Static and Dynamic Features Through Transfer Learning
Title:
Fast and Efficient Malware Detection with Joint Static and Dynamic Features Through Transfer Learning
Journal Title:
Applied Cryptography and Network Security
Keywords:
Publication Date:
28 May 2023
Citation:
Ngo, M. V., Truong-Huu, T., Rabadi, D., Loo, J. Y., & Teo, S. G. (2023). Fast and Efficient Malware Detection with Joint Static and Dynamic Features Through Transfer Learning. Lecture Notes in Computer Science, 503–531. https://doi.org/10.1007/978-3-031-33488-7_19
Abstract:
In malware detection, dynamic analysis extracts the runtime be- havior of malware samples in a controlled environment and static analysis extracts features using reverse engineering tools. While the former faces the challenges of anti-virtualization and evasive behavior of malware samples, the latter faces the challenges of code obfuscation. To tackle these drawbacks, prior works proposed to develop detection models by aggregating dynamic and static fea- tures, thus leveraging the advantages of both approaches. However, simply concatenating dynamic and static features raises an issue of imbalanced contribution due to heterogeneous dimensions of feature vectors, resulting in not much performance improvement. Yet, dynamic analysis is a time-consuming task and requires a se- cure environment, leading to detection delays and high costs for maintaining the analysis infrastructure. In this paper, we first in- troduce a novel method of constructing aggregated features via concatenating latent features learned through deep learning with equally-contributed dimensions. We then develop a knowledge dis- tillation technique to transfer knowledge learned from aggregated features by a teacher model to a student model trained only on static features. We carry out extensive experiments with a dataset of 86 709 samples including both benign and malware samples. The experimental results show that the teacher model trained on aggre- gated features constructed by our method outperforms the state- of-the-art models with an improvement of up to 2.38% in detection accuracy. The distilled student model not only achieves high per- formance (97.81% in terms of accuracy) as that of the teacher model but also significantly reduces the detection time (from 70 046.6ms to 194.9ms) without requiring dynamic analysis.
License type:
Publisher Copyright
Funding Info:
There was no specific funding for the research done
Description:
This version of the article has been accepted for publication, after peer reviewand is subject to Springer Nature’s AM terms of use, but is not the Version of Record and does not reflect post-acceptance improvements, or any corrections. The Version of Record is available online at: http://dx.doi.org/10.1007/978-3-031-33488-7_19
ISSN:
9783031334887
ISBN:
9783031334870
Files uploaded:

File Size Format Action
978-3-031-33488-7-19-1.pdf 1.92 MB PDF Open