Skip to Main Content
 

Global Search Box

 
 
 
 

Files

File List


revised-thesis-prabesh-pathak.pdf (2.89 MB)

ETD Abstract Container

Abstract Header

Leveraging attention-based deep neural networks for security vetting of Android applications

Pathak, Prabesh
http://rave.ohiolink.edu/etdc/view?acc_num=bgsu1617215079338328

Abstract Details

2021, Master of Science (MS), Bowling Green State University, Computer Science.
Traditional machine learning and deep learning algorithms work as a black box and lack explainability. Attention-based mechanisms come to the rescue to address interpretability of such models by providing insights and explaining why a model makes its decisions. In recent years, attention-based mechanisms have been able to explain the role of the inputs in natural language processing. This success motivates us to extend these attention-based approaches for security vetting of Android apps. An Android app's code contains API calls that can give us some clue of if the app is either malicious or benign. By observing the pattern of such API calls being invoked, we can build a model to separate benign apps from malicious apps. In this thesis work, using attention-based mechanisms, we aim to find the important API calls that are responsible for reflecting the maliciousness of android apps. As an example, we target to identify a subset of APIs that malicious apps exploit, which might help the community discover a new signature of malware. In our experiment, we work with two attention models: Bi-LSTM Attention and Self-Attention. Our classification models achieve an area under the precision-recall curve (AUC) score of 0.9391 and 0.9074 with Bi-LSTM Attention and Self-Attention, respectively. Using the attention weights from our models, we are also able to extract top 200 API-calls from each of our models that reflect the malicious behavior of the apps.
Sankardas Roy (Advisor)
Robert C. Green II (Committee Member)
Qing Tian (Committee Member)
71 p.
Computer Science
Android; Android Security; Attention; Classification; Deep Learning

Recommended Citations

Citations

  • Pathak, P. (2021). Leveraging attention-based deep neural networks for security vetting of Android applications [Master's thesis, Bowling Green State University]. OhioLINK Electronic Theses and Dissertations Center. http://rave.ohiolink.edu/etdc/view?acc_num=bgsu1617215079338328

    APA Style (7th edition)

  • Pathak, Prabesh. Leveraging attention-based deep neural networks for security vetting of Android applications. 2021. Bowling Green State University, Master's thesis. OhioLINK Electronic Theses and Dissertations Center, http://rave.ohiolink.edu/etdc/view?acc_num=bgsu1617215079338328.

    MLA Style (8th edition)

  • Pathak, Prabesh. "Leveraging attention-based deep neural networks for security vetting of Android applications." Master's thesis, Bowling Green State University, 2021. http://rave.ohiolink.edu/etdc/view?acc_num=bgsu1617215079338328

    Chicago Manual of Style (17th edition)

bgsu1617215079338328
489
© 2021, all rights reserved.
This open access ETD is published by Bowling Green State University and OhioLINK.