An Evaluation of Machine Learning Approaches for Android Malware Detection to Enhance Cyber Security

Android has been the go-to OS for smartphones for the last several years. A serious privacy and security risk is malicious assaults on Android. Thus, methods for detecting malware in Android apps are important. Android malware detection has become crucial in enhancing cybersecurity, given the increasing prevalence of malicious applications targeting mobile platforms. ML methods such as SVR, NB, RF, LSTM, and SVM are tested on the Android Malware Detection Dataset by Kaggle in this research. The dataset, consisting of features extracted by Android applications (APKs), was preprocessed by handling missing values, label encoding, and standard scaling to optimise data quality and model performance. Among the evaluated models, SVR with an RBF kernel achieved the highest accuracy of $95.74 \%$ and demonstrated strong performance across precision ($94.76 \%$), recall ($98.06 \%$), and F1-score $(96.38 \%)$. Comparative analysis shows that while Naïve Bayes struggles with complex data, models like Random Forest and LSTM also achieved high accuracy ($90.93 \%$ and $94.59 \%$, respectively) with strong recall, making them suitable alternatives. The findings suggest that SVR, LSTM, and Random Forest are particularly effective for Android malware detection, providing a robust foundation for enhancing mobile cybersecurity.