Thesis
UNCERTAINTY-AWARE MACHINE LEARNING TECHNIQUES FOR SUSTAINABLE MALWARE DETECTION
Washington State University
Master of Science (MS), Washington State University
12/2024
DOI:
https://doi.org/10.7273/000007182
Abstract
Malware remains a significant global cybersecurity threat, with millions of new variants appearing rapidly, highlighting the urgent need for effective automated detection methods. Traditional malware detection systems face challenges in adapting to the rapidly evolving landscape, particularly due to concept drift, where malware characteristics change over time, leading to performance degradation. This thesis introduces an innovative framework for uncertainty-aware malware detection using Gaussian processes (GPs) to quantify prediction uncertainty and enhance the reliability of malware detection.
Our approach integrates GPs into machine learning models to address the limitations of existing malware detection techniques. By providing a probabilistic measure of confidence in predictions, it enables selective abstention from highly uncertain classification decisions, reducing false positives and negatives. We applied this framework to two qualitatively different mobile malware datasets.
Through comprehensive evaluations, we demonstrate that our GP-based models significantly outperform traditional methods in accuracy and adaptability to evolving malware in
Android. Extensive ablation studies validated our hypothesis on the effectiveness of uncertainty quantification in improving the overall detection performance. Our results provide valuable insights into malware behavior dynamics and the importance of adaptive detection strategies.
Overall, this research contributes significantly to the development of robust and adaptable mobile malware detection solutions by integrating uncertainty quantification into machine learning-based detection systems. This study underscores the potential for GPs to enhance long-term performance in real-world applications, addressing the critical need for reliable detection mechanisms in an increasingly complex cyberthreat landscape.
Metrics
13 File views/ downloads
3 Record Views
Details
- Title
- UNCERTAINTY-AWARE MACHINE LEARNING TECHNIQUES FOR SUSTAINABLE MALWARE DETECTION
- Creators
- Dheeraj Vurukuti
- Contributors
- Janardhan Rao Doppa (Co-Chair)Haipeng Cai (Co-Chair)Yan Yan (Committee Member)
- Awarding Institution
- Washington State University
- Academic Unit
- School of Electrical Engineering and Computer Science
- Theses and Dissertations
- Master of Science (MS), Washington State University
- Publisher
- Washington State University
- Number of pages
- 81
- Identifiers
- 99901195201401842
- Language
- English
- Resource Type
- Thesis