Deep Learning Based Malware Detection Tool Development for Android Operating System
<p><em>In today's world that called technology age, smartphones have become indispensable for users in many areas such as internet usage, social media usage, bank transactions, e-mail, as well as communication. The Android operating system is the most popular operating system that u...
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
EduSoft publishing
2021-12-01
|
| Series: | Brain: Broad Research in Artificial Intelligence and Neuroscience |
| Subjects: | |
| Online Access: | https://www.edusoft.ro/brain/index.php/brain/article/view/1186 |
| _version_ | 1797335713502461952 |
|---|---|
| author | Mahmut TOKMAK Ecir Uğur KÜÇÜKSİLLE Utku KÖSE |
| author_facet | Mahmut TOKMAK Ecir Uğur KÜÇÜKSİLLE Utku KÖSE |
| author_sort | Mahmut TOKMAK |
| collection | DOAJ |
| description | <p><em>In today's world that called technology age, smartphones have become indispensable for users in many areas such as internet usage, social media usage, bank transactions, e-mail, as well as communication. The Android operating system is the most popular operating system that used with a rate of 85.4% in smartphones and tablets. Such a popular and widely used platform has become the target of malware. Malicious software can cause both material and moral damages to users.</em></p><p><em>In this study, malwares that targeting smart phones were detected by using static, dynamic and hybrid analysis methods. In the static analysis, feature extraction was made in 9 different categories. These attributes are categorized under the titles of requested permissions, intents, Android components, Android application calls, used permissions, unused permissions, suspicious Android application calls, system commands, internet addresses. The obtained features were subjected to dimension reduction with principal component analysis and used as input to the deep neural network model. With the established model, 99.38% accuracy rate, 99.36% F1 score, 99.32% precision and 99.39% sensitivity values were obtained in the test data set.</em></p><p><em>In the dynamic analysis part of the study, applications were run on a virtual smartphone, and Android application calls with strategic importance were obtained by hooking. The method called hybrid analysis was applied by combining the dynamically obtained features with the static features belonging to the same applications. With the established model, 96.94% accuracy rate, 96.78% F1 score, 96.99% precision and 96.59% sensitivity values were obtained in the test data set.</em></p> |
| first_indexed | 2024-03-08T08:42:22Z |
| format | Article |
| id | doaj.art-7314d167342c4e9c98e109e4558b1708 |
| institution | Directory Open Access Journal |
| issn | 2067-3957 |
| language | English |
| last_indexed | 2024-03-08T08:42:22Z |
| publishDate | 2021-12-01 |
| publisher | EduSoft publishing |
| record_format | Article |
| series | Brain: Broad Research in Artificial Intelligence and Neuroscience |
| spelling | doaj.art-7314d167342c4e9c98e109e4558b17082024-02-01T18:00:41ZengEduSoft publishingBrain: Broad Research in Artificial Intelligence and Neuroscience2067-39572021-12-011242856992Deep Learning Based Malware Detection Tool Development for Android Operating SystemMahmut TOKMAK0Ecir Uğur KÜÇÜKSİLLE1Utku KÖSE2Isparta University of Applied Sciences, Gelendost Vocational School, Isparta, TurkeySuleyman Demirel University, Department of Computer Engineering, Isparta, TurkeySuleyman Demirel University, Department of Computer Engineering, Isparta, Turkey<p><em>In today's world that called technology age, smartphones have become indispensable for users in many areas such as internet usage, social media usage, bank transactions, e-mail, as well as communication. The Android operating system is the most popular operating system that used with a rate of 85.4% in smartphones and tablets. Such a popular and widely used platform has become the target of malware. Malicious software can cause both material and moral damages to users.</em></p><p><em>In this study, malwares that targeting smart phones were detected by using static, dynamic and hybrid analysis methods. In the static analysis, feature extraction was made in 9 different categories. These attributes are categorized under the titles of requested permissions, intents, Android components, Android application calls, used permissions, unused permissions, suspicious Android application calls, system commands, internet addresses. The obtained features were subjected to dimension reduction with principal component analysis and used as input to the deep neural network model. With the established model, 99.38% accuracy rate, 99.36% F1 score, 99.32% precision and 99.39% sensitivity values were obtained in the test data set.</em></p><p><em>In the dynamic analysis part of the study, applications were run on a virtual smartphone, and Android application calls with strategic importance were obtained by hooking. The method called hybrid analysis was applied by combining the dynamically obtained features with the static features belonging to the same applications. With the established model, 96.94% accuracy rate, 96.78% F1 score, 96.99% precision and 96.59% sensitivity values were obtained in the test data set.</em></p>https://www.edusoft.ro/brain/index.php/brain/article/view/1186android malware analysis, static analysis, dynamic analysis, hybrid analysis, deep learning |
| spellingShingle | Mahmut TOKMAK Ecir Uğur KÜÇÜKSİLLE Utku KÖSE Deep Learning Based Malware Detection Tool Development for Android Operating System Brain: Broad Research in Artificial Intelligence and Neuroscience android malware analysis, static analysis, dynamic analysis, hybrid analysis, deep learning |
| title | Deep Learning Based Malware Detection Tool Development for Android Operating System |
| title_full | Deep Learning Based Malware Detection Tool Development for Android Operating System |
| title_fullStr | Deep Learning Based Malware Detection Tool Development for Android Operating System |
| title_full_unstemmed | Deep Learning Based Malware Detection Tool Development for Android Operating System |
| title_short | Deep Learning Based Malware Detection Tool Development for Android Operating System |
| title_sort | deep learning based malware detection tool development for android operating system |
| topic | android malware analysis, static analysis, dynamic analysis, hybrid analysis, deep learning |
| url | https://www.edusoft.ro/brain/index.php/brain/article/view/1186 |
| work_keys_str_mv | AT mahmuttokmak deeplearningbasedmalwaredetectiontooldevelopmentforandroidoperatingsystem AT ecirugurkucuksille deeplearningbasedmalwaredetectiontooldevelopmentforandroidoperatingsystem AT utkukose deeplearningbasedmalwaredetectiontooldevelopmentforandroidoperatingsystem |