**6. Conclusions**

Smartphones are becoming more and more popular, constituting a profitable target for hackers due to their susceptibility to security breaches. Android is an open gate for attackers who exploit it with malicious applications, benefiting from the system's security flaws. An emerging method for signature-based malicious attack detection is the antivirus applications against new malware, created with AI, machine learning, and deep learning algorithms that predict malware. In this study, a security system was built and designed based on the support vector machine (SVM), k-nearest neighbors (KNN), linear discriminant analysis (LDA), long short-term memory (LSTM), convolution neural network-long short-term memory (CNN-LSTM), and autoencoder algorithms. According to the promising results of the present research, the following conclusions can be drawn:

The proposed system was evaluated and examined using two standard Android malware applications datasets: CICAndMal2017 and Drebin. The SVM, KNN, and LDA methods proved to be efficient machine learning algorithms and successfully detected malware, with SVM being the most effective. The LSTM and CNN-LSTM models are proposed to detect malicious applications, with the LSTM model being more efficient for developing Android security. Sensitive analysis examining the metrics MSE, RMSE, and R<sup>2</sup> revealed the errors between the predicted output and the target values in the validation phase. The LSTM and CNN-LSTM algorithms achieved fewer prediction errors in the Drebin dataset, while the SVM method was more effective in the case of the CICAndMal2017 dataset. The validation phase results of the machine learning and deep learning methods were satisfying, with the LSTM and SVM models achieving superior performance. The results of the present study were compared with recent research findings, confirming the robustness and effectiveness of our results. We implemented machine learning and deep learning algorithms and experimented with them to obtain optimal malware detection. Both of the proposed classifiers achieved good accuracy, but the LSTM accuracy was 99.40%, indicating it can outperform other state-of-the-art models.

**Author Contributions:** Conceptualization, T.H.H.A. and H.A.; methodology, T.H.H.A.; software, T.H.H.A.; validation, T.H.H.A. and H.A.; formal analysis, T.H.H.A. and H.A.; investigation, T.H.H.A. and H.A.; resources, T.H.H.A.; data curation, T.H.H.A. and H.A.; writing—original draft preparation, T.H.H.A. and H.A.; writing—review and editing, H.A.; visualization, T.H.H.A. and H.A.; supervision, T.H.H.A.; project administration, T.H.H.A. and H.A.; funding acquisition, T.H.H.A. and H.A. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research and the APC were funded by the Deanship of Scientific Research at King Faisal University for financial support under gran<sup>t</sup> No. NA00036.

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** The data presented in this study are available here: https://www. kaggle.com/saurabhshahane/Android-permission-dataset; https://www.kaggle.com/shashwatwork/ android-malware-dataset-for-machine-learning (accessed on 25 Novmber 2021).

**Acknowledgments:** The authors extend their appreciation to the Deanship of Scientific Research at King Faisal University for funding this research work through project number NA00038.

**Conflicts of Interest:** The authors declare that they have no conflict of interest.
