Özet:

Today, smartphones have become an integral part of human life. Android operating system has a high rate of use among these devices. Thanks to its advanced features, it allows users to store their personal information such as photos, health data, identity information and bank information. It is a very targeted operating system by malware developers due to its widespread use and advanced features. In this study, the success in the detection of Android malicious software applications was first extracted using the profound autocoding architecture. In the next phase, the machine learning methods were classified using the algorithms of Rasgele Forest (RO), K-The Nearest Neighbor (K-EYK) and Decision Tree (KA). Experimental results showed that character extraction increased success using deep auto coder and basic component analysis. According to the analysis, the random forest algorithm has 94.40% accuracy.

Özet:

Nowadays, smart phones have become an indispensable part of human life. The Android operating system has the highest utilization rate among these devices. Its advanced features allow users to store personal information such as photos, health data, identity information, and bank information. It is the most targeted operating system by malware developers because of its widespread usage and advanced features. In this study, in order to increase the success in the detection of Android malware applications, feature extraction was performed by using deep autoencoders. In the next step, data are classified by Random Forest (RF), K-Nearest Neighbor (K-NN) algorithm and Decision Tree (DT) algorithms. Experimental results showed that feature extraction using deep autoencoders and principal component analysis increased the accuracy of model. According to the analyses made, it has been observed that Random Forest algorithm had the best accuracy with 94.40%.

Anahtar Kelimeler: