Classifying IoT Malware with Limited Data: A Few-Shot Learning Framework

Abstract

The rapid growth of Internet of Things (IoT) devices has amplified malware risks, challenging traditional detection methods. Conventional machine learning relies on large labeled datasets, which often fall short in addressing emerging malware variants. Furthermore, the uneven distribution of IoT malware families across different CPU architectures complicates the effectiveness of detection methods. To overcome these challenges, we propose leveraging few-shot learning (FSL) for IoT malware analysis. This innovative approach enables accurate detection with limited sample sizes, improves adaptability to new and evolving threats, and reduces the burden of data labelling. Although FSL holds great promise, it remains largely underexplored in the domain of IoT malware detection. This research investigates the capability of FSL to enhance classification accuracy and robustness, particularly in cases with scarce or imbalanced datasets. Our findings aim to contribute to the development of adaptable and efficient solutions for the ever-evolving threat landscape in IoT environments.

DOI: https://doi.org/24135/ICONIP2