Harnessing Transfer Learning for Rapid Malware Family Classification in Large-Scale Cyber Datasets
Rishabh Mohan, Himanshu Gupta · EPJ Web of Conferences · 2025
Malware classification remains a critical challenge in cybersecurity, particularly in an era of rapidly evolving threats and vast datasets. Traditional machine learning methods struggle to keep pace with the diversity and volume of malware samples. This paper explores the application of transfer learning to classify malware families efficiently in large-scale cyber datasets. Leveraging pre-trained deep learning models, we demonstrate significant improvements in classification accuracy and speed while reducing the dependency on extensive labelled data. By utilizing pre-trained architectures such as Convolutional Neural Networks (CNNs) and Transformers, we exploit their ability to learn transferable features, minimizing the need for domain-specific knowledge. Furthermore, our methodology incorporates fine-tuning and domain adaptation techniques to ensure relevance and robustness in malware classification tasks. We conduct extensive experiments using real-world malware datasets, showcasing that transfer learning not only reduces computational overhead but also achieves superior performance compared to traditional approaches. Our results demonstrate the efficacy of this approach in handling skewed data sets, preventing overfitting, and enabling rapid deployment in dynamic cyber spaces. Our research results emphasize the importance of transfer learning in making cybersecurity solutions adaptable and effective in the face of evolving malware threats.