Adware Malware Detection by Using Random Forest and Naive Bayes Classification with Information Gain

The presence of adware in mobile applications poses a significant threat to user privacy and security. This research aims to study the effective adware malware detection by using Random Forest and Naive Bayes classification with Information Gain. Manifest permission-based features are utilized as indicators for identifying adware applications. The CCCS-CIC-AndMal-2020 dataset is employed for training and evaluating the proposed model. Through feature selection techniques, the top 10 most informative features are identified, which play a crucial role in adware detection. The performance of the model is assessed using various evaluation metrics, including Accuracy, Precision, True Positive, and True Negative. The experimental results demonstrate the effectiveness of the proposed approach in detecting adware, providing valuable insights for the development of robust and reliable adware detection systems. The Random Forest algorithm demonstrated superior performance compared to Naive Bayes in detecting adware, exhibiting higher values for all performance metrics, suggesting its effectiveness as a classifier for identifying adware instances with the accuracy of 82.16%, Precision of 89.80%, True Positive Rate of 68.00%, and True Negative Rate 93.47%.