A Multi-Feature User Authentication Model Based on Mobile App Interactions

Yosef Ashibani,Qusay H Mahmoud

doi:10.1109/access.2020.2996233

Abstract

Knowledge-based authentication approaches such as the use of passwords and personal identification numbers (PINs) are the most common ways of authenticating users. The main problem with such approach is relying on simple authentication login credentials at the login stage, and assuming the user is still the same between access sessions makes applications and networks vulnerable to access by unauthorized users. Application-level access patterns on smartphone and tablet devices can be utilized to provide an approach for continuously authenticating and identifying users. This paper presents a user authentication and identification method based on mobile application access patterns, and throughout the paper we use a smart home environment as a motivating scenario. To enhance the classification process, many features have been extracted and utilized which considerably improved differentiating between users and eliminating similarities in the access usage patterns. The proposed model has been evaluated using two datasets, and the results show an ability to authenticate users with high accuracy in terms of low false positive, false negative, and equal error rates. Overall, the statistical analysis of the extracted multi-features and the results show that the feasibility of decision-making based on app interactions can lead to high accuracy.

Highlights

A smart home can be defined as a home equipped with connected Internet-of-Things (IoT) devices that can be remotely accessed and controlled
EXPERIMENTAL EVALUATION AND RESULTS To evaluate the performance of the presented method, the datasets UbiqLog4UCI [40] and LiveLab [41] collected from real users is utilized, and the identification performance is considered as the accuracy metric when classifying an access session to one of the enrolled users
The first classifier is the random forest (RF) classifier, which fits a number of decision tree classifiers on various subsamples of instances and utilizes

Summary

Introduction

A smart home can be defined as a home equipped with connected Internet-of-Things (IoT) devices that can be remotely accessed and controlled. In addition to accessing, operating, and controlling home appliances, smart home networks provide many other services to home residents, such as entertainment storage information and personal files. Wink [1], Samsung’s SmartThings and HomeKit [2] are smart home platforms. These home platforms are built based on the cloud backend service where control management and authentication are performed mostly through an installed application on end-user devices, such as smartphones and tablets. Access to smart home services is mostly achieved remotely through the users’ end-devices which have become essential tools for accessing and operating smart home networks. Smart home systems conveniently provide services to home residents, there are many security issues that need to be considered.

Objectives

Results

Conclusion