Research Article
A Generalized Unknown Malware Classification
@INPROCEEDINGS{10.1007/978-3-031-25538-0_41, author={Nanda Rani and Ayushi Mishra and Rahul Kumar and Sarbajit Ghosh and Sandeep K. Shukla and Priyanka Bagade}, title={A Generalized Unknown Malware Classification}, proceedings={Security and Privacy in Communication Networks. 18th EAI International Conference, SecureComm 2022, Virtual Event, October 2022, Proceedings}, proceedings_a={SECURECOMM}, year={2023}, month={2}, keywords={Malware classification Deep learning Cyber Security Malware}, doi={10.1007/978-3-031-25538-0_41} }- Nanda Rani
Ayushi Mishra
Rahul Kumar
Sarbajit Ghosh
Sandeep K. Shukla
Priyanka Bagade
Year: 2023
A Generalized Unknown Malware Classification
SECURECOMM
Springer
DOI: 10.1007/978-3-031-25538-0_41
Abstract
Although state-of-the-art image-based malware classification models give the best performance, these models fail to consider real-world deployment challenges due to various reasons. We address three such problems through this work: limited dataset problems, imbalanced dataset problems, and lack of model generalizability. We employ a prototypical network-based few-shot learning method for a limited dataset problem and achieve 98.71% accuracy while training with only four malware samples of each class. To address the imbalanced dataset problem, we propose a class-weight technique to increase the weightage of minority classes during the training. The model performs well by improving precision and recall from 0% to close to 60% for the minority class. For the generalized model, we present a meta-learning-based approach and improve model performance from 48% to 72.06% accuracy. We report performances on five diverse datasets. The proposed solutions have the potential to set benchmark performance for their corresponding problem statements.
