Android Malware Detection Based on Grammatical Evaluation Algorithm and XGBoost

Smartphones are prevalent in the modern digital era and typically work on open-source software, which makes it easier for malicious code to enter the system. Android malware infiltrates the smartphone and corrupts the user’s sensitive and personal data. So Multiple methods have been developed to reduce the risks presented by malware. Signature-based detection is the most common method for detecting Android malware. This procedure cannot detect unknown malicious software, which is considered a disadvantage. Due to this issue, Machine Learning (ML) has been implemented to detect malware threats. The objective of ML algorithms is to improve classification precision. Due to the complexity of actual datasets,, conventional classifiers are ineffective at classifying malicious applications. The security sector has begun to identify hybrid machine-learning systems. This paper proposes a hybrid system for Android smartphone malware detection. We developed a detection model using Extreme Gradient Boosting (XGBoost) and Grammatical Evaluation (GE) to determine the optimal parameters. Three well-known datasets in the field of Android Malware are used to apply the experimental results. Compared to the results derived through conventional parameter adjustment, the performance of the proposed model was superior. Our method attained an accuracy of 98% for CICMalDroid-2020, 99.02% for Drebin, and 99.28% for Malgenome, according to the results obtained.