Date of Award
August 2026
Document Type
Thesis
Degree Name
Master of Science (MS)
Department
Electrical Engineering
First Advisor
Naima Kaabouch
Abstract
Internet of Things devices face a growing volume of malware, and detection systems run into two practical problems: too little labeled data exists for a threat when it first appears, and malware families often resemble each other too closely for flat feature representations to separate. Both problems were addressed on the Ransomware Dataset 2024 using two purpose-built architectures. Detection paired an autoencoder with a meta-learning classifier that adapted from a single labeled example, reached 87.6% accuracy at one-shot, and outperformed standard baselines through fifty-shot. Classification used a state-space sequence model that captures relationships between features a flat vector misses, and reached an 89% macroF1-score on structurally similar ransomware families, 14 points above the next-best baseline. These results show that reliable detection and classification remain achievable even under the data scarcity and family-level similarity that leave many IoT devices exposed to malware they have never encountered before
Recommended Citation
Andam, Emmanuela, "Malware Detection and Behavioral Analysis in IoT Environments: A Multi-Study Deep Learning Approach" (2026). Theses and Dissertations. 13863.
https://commons.und.edu/theses/13863