Investigation of users' mobile application security awareness using machine learning techniques

The enriched functionality and interaction features of mobile phones, such as ubiquity, ease of instant connection, application diversity, personalization, flexibility, distribution and location based services, have made them the world's first means of communication. Mobile phones have now become an indispensable element for most people. Users interact in the internet world with the applications installed on these devices. Mobile application stores offer users the opportunity to discover and download thousands of applications in various categories. Every day, millions of people use these stores to find applications that suit their needs or interests. By downloading applications from these stores, users have the opportunity to communicate, have fun, obtain information, shop, make financial transactions, travel planning and many other transactions. In this study, a research survey data set of the Dataverse platform belonging to Harvard University was used. The survey was conducted with 10,208 people in more than 15 countries. The data set includes information such as demographic characteristics, educational information and mobile application usage behavior of the survey participants. The main goal of this study is to analyze the profiles of mobile device users and their application usage purposes and needs, and to determine the factors that affect users' decisions to choose, use and abandon an application, using machine learning techniques. In the research, the stages of finding, selecting and abandoning the application were tested in the analysis performed on the data set with Logistic Regression (LR), Random Forest (RF), Support vector machine (SVM), K-Nearest Neighbors (KNN) machine learning algorithms. During the testing stages, Accuracy, Precision, Recall and F1-Score (F-Measure) values were examined. "User demographic characteristics affect the behavior of finding the application." In verifying the 1st Hypothesis established as follows, the SVM machine learning algorithm was the most successful algorithm with an accuracy rate of 0.930 and an F1-Score value of 0.950. In confirming the 2nd Hypothesis of the research, which was established as "The demographic characteristics of the user affect the application selection behavior", the SVM machine learning algorithm was the most successful algorithm with an accuracy rate of 0.920 and an F1-Score value of 0.950. "User demographic characteristics affect application abandonment behavior." In the 3rd Hypothesis, according to the SVM machine learning algorithm, the accuracy rate was 0.940 and the F1 Score value was 0.970, making it the most successful algorithm. According to the analysis results obtained, it has been seen that the demographic characteristics of mobile application users have an impact on the behavior of finding, choosing and leaving the application and can be accurately predicted by machine learning methods. "There is a significant relationship between mobile application users' behavior in finding and choosing the application." Pearson Correlation model was used to test Hypothesis 4. According to the results of this model, the Average Pearson Correlation Coefficient was 0.215. "There is a significant relationship between the application selection and abandonment behavior of mobile application users." Pearson Correlation model was used to test Hypothesis 5. According to the results of this model, the Average Pearson Correlation Coefficient was 0.230. According to these obtained values, it was seen that there was a weak positive relationship between the user's behavior of finding, selecting and leaving the mobile application. It is thought that this research will help users increase their awareness of mobile application security and privacy and make informed choices. It also provides valuable data to mobile app developers to better understand users' needs and improve the app experience.