Behavior-Based Security for Mobile Devices Using Machine Learning Techniques
Download as PPTX, PDF0 likes56 views
This research project aims to enhance mobile device security through a behavior-based mobile application leveraging machine learning techniques for anomaly detection. The developed Android application collects user behavior data to identify unique users, which could aid in security, service recommendations, and offline behavior modeling. Initial results indicate that the proposed methods effectively model user behavior with high accuracy and potential for varied applications.
![Figure 2. Basic Relief Algorithm [12], [13]](https://image.slidesharecdn.com/ijaia-1-181122090428/85/Behavior-Based-Security-for-Mobile-Devices-Using-Machine-Learning-Techniques-6-320.jpg)
Behavior-Based Security for Mobile Devices Using Machine Learning Techniques
- 1. Sherif Rashad1 and Jonathan Byrd2,3, 1University of Southern Indiana, USA, 2Colorado State University, USA and 3Morehead State University, USA.
- 2. ABSTRACT The goal of this research project is to design and implement a mobile application and machine learning techniques to solve problems related to the security of mobile devices. We introduce in this paper a behavior-based approach that can be applied in a mobile environment to capture and learn the behavior of mobile users. The proposed system was tested using Android OS and the initial experimental results show that the proposed technique is promising, and it can be used effectively to solve the problem of anomaly detection in mobile devices. KEYWORDS Behavior-based Security, Anomaly Detection, Mobile Profiles, Mobile Computing.
- 3. INTRODUCTION In this research project we explore the rapidly evolving capabilities of modern smartphones with a goal to design an innovative approach to improve the security of smartphones and better understand the dynamic behavior of mobile users. With an increasing number of smartphone users, it becomes highly important to improve the efficiency and security of these smart phones. Individual computer users interact with machine interfaces in different ways, which includes analyzing keystroke patterns, CLI commands, or navigation pathways through GUIs. We can often differentiate between individual users based on these interactions. In this project, we explore the problem of distinguishing Android phone users based on their behavior using machine learning techniques. We have developed an Android application that collects different features that represent user behavior information such as battery life, number of applications running, tilt of the device, location, etc, and use this data to learn the behavior of mobile users. This approach has a potential application in many areas including security. Anomaly detection techniques based on user behavior model could be used to detect malicious users who have gained access to a device. Additionally, it could be used to detect unusual lowlevel commands run by malware. In addition to providing interesting information, such as which features are most salient, classifying unique users has potential real-world applications such as:
- 4. •Distinguishing between users on multi-user systems. If a single device is shared by multiple users, an application could prevent access of information from the incorrect user, or automatically log into accounts based on the user behavior. • Service/product recommendations • Building offline behavior models to find correlations between demographics (age, gender, location, etc.) and user behaviour .
- 5. Figure. 1 Basic Architecture of the Proposed Approach
- 6. Figure 2. Basic Relief Algorithm [12], [13]
- 7. Figure 3. The Interface of the Developed Android Collector App
- 8. CONCLUSION The goal of this research project is to propose a smart approach for the problem of behavior-based security of mobile devices using machine learning techniques. We focused on the problem of anomaly detection, but the proposed approach has possible applications in many other areas to identify the mobile users by learning the behavior of using the mobile devices. An Android application was used to collect location and behavior data of mobile users to build a behavior-based security system. The Relieff feature ranking algorithm was used effectively to reduce the dimension and the volume of the collected data. Different algorithms were used to analyze the behavior data of mobile users. Our initial experimental results show that the proposed techniques can be used to model the behavior of mobile users and identify the mobile users based on behavior data with a high accuracy and a reduced set of collected data. Future work includes improving the overall performance and considering the most relevant features based on the mobile user. We need to test multiple users on the same device and we expect that would provide useful data. We also need to consider more users and other possible applications of the proposed techniques to provide new identity-based applications and services.
- 9. International Journal of Artificial Intelligence and Applications (IJAIA) http://www.airccse.org/journal/ijaia/ijaia.html For More Details : http://aircconline.com/ijaia/V9N4/9418ijaia01.pdf