Research Article
CAFM: Precise Classification for Android Family Malware
@INPROCEEDINGS{10.1007/978-3-030-96791-8_28, author={Dan Li and Runbang Pan and Ning Lu and Wenbo Shi}, title={CAFM: Precise Classification for Android Family Malware}, proceedings={Security and Privacy in New Computing Environments. 4th EAI International Conference, SPNCE 2021, Virtual Event, December 10-11, 2021, Proceedings}, proceedings_a={SPNCE}, year={2022}, month={3}, keywords={Android malware Family Classification}, doi={10.1007/978-3-030-96791-8_28} }- Dan Li
Runbang Pan
Ning Lu
Wenbo Shi
Year: 2022
CAFM: Precise Classification for Android Family Malware
SPNCE
Springer
DOI: 10.1007/978-3-030-96791-8_28
Abstract
Family malware classification is becoming progressively urgent because of the increasing diversity of family malware and the different hazards it causes. There is a growing concern that classification is at a disadvantage owing to its problems. For one thing, obtaining the crucial features of innumerable families is arduous. For another, constructing a classification model that fully learns multi-class samples is intricate. To solve these problems, it proposes a precise classification for Android family malware called CAFM in this paper. It profoundly analyzes the relationship between the information implicit in features and the degree of differentiation among families. We select the features containing context information as feature representations. In addition, it employs a specially designed deep neural network model with upgraded learning capability for grasping the continuous features of family malware utterly. Experimental verification on a real-world dataset shows that the CAFM can effectively implement family classification, and the classification accuracy reaches 97.73% when the length of the opcode sequence is 700. Compared with other classifiers, the Kappa coefficient of the comprehensive evaluation indicator also reached 0.9725 and is at least 0.1225 higher than comparison classifiers.
