1. Detection of malware in Android environment using machine learning techniques.
- Author
-
Salman, Farah Rafid and Abdul Rahman, Amer Abdulmajeed
- Subjects
- *
DATA scrubbing , *FEATURE selection , *APPLICATION stores , *RANDOM forest algorithms , *ERROR rates - Abstract
These days, malicious apps represent a serious threat to Android consumers, developers, and app stores. Because malware has become more complicated, it has undergone constant modification, and its attacks have become more destructive, researchers are working to develop new techniques for detecting it. Having a clean and balanced dataset is one of the biggest obstacles in malware detection. In this research, a detection approach to detect malware proposed to increase precision and decrease error rates by preprocessing and balancing the dataset. To attain these purposes, feature selection using C4.5 method was used to choose the most important and relevant features and the inefficient features are eliminated. The Synthetic Minority Oversampling Technique (SMOTE) approach used in the suggested method to further balance the dataset. Then, the Random Forest (RF) classifier was used to create the detection model. The data cleaning techniques and the use of SMOTE enhance the model performance. The obtained results indicate that the criteria of Precision, Recall, Accuracy and F1 score are high. The proposed methodology demonstrates efficacy in the detection of both pre-existing malware within the dataset and newly emerging malware. The accuracy for unbalanced dataset using RF classifier where (94.79,93.28,97.06,96.66) for (all feature, 10 features, 20 features, 40 features) respectively and the accuracy for balanced dataset when classified using RF classifier where (97.75,94.27,98.83,98.19) for (all feature, 10 features, 20 features, 40 features) respectively, CIC-AndMal2017 dataset used to in this research. [ABSTRACT FROM AUTHOR]
- Published
- 2024
- Full Text
- View/download PDF