CiAFP: Category-based Classification of iOS Apps by mining frequent permissions
Neetu Sardana, Arpita Jadhav Bhatt · Proceedings of the 2022 Fourteenth International Conference on Contemporary Computing · 2022
With the presence of 1.96 million apps, on the App Store, iOS is one of the most prevalently used mobile operating systems in terms of its users and after its counterpart Android. These apps have penetrated every aspect of our lives. They are used in payments, e-shopping, navigation, instant messaging and are also integrated with IoT devices to transmit data worldwide. These apps provide great convenience however they raise privacy and security concerns since a lot of users’ personal information can be accessed by the apps. Although Apple adopts a permission-based access control mechanism to confine apps from retrieving users’ personal information like address book, geo- coordinates, and photo gallery, the user still faces a significant menace of privacy leakage due to over-privileged permission, which means extra permissions affirmed by an app but then has nothing to do with its functionality. Unfortunately, the unavailability of tools to detect over-privilege permissions compels the user to grant all permissions declared by an app intensifies the risk of a privacy breach. Previous studies to detect over-privileged apps have been conducted for Android but minimal work has been done for iOS. To combat this problem, we have proposed a framework CiAFP that employs Association Rule Mining (ARM) technique to mine frequent permission patterns of iOS apps. This helps to determine over-privilege permissions. We have mined permission patterns for 331 iOS apps from 3 different categories namely Education, Finance, and Health & Fitness, and found that these categories have 40%,26%, and 24% over-privileged permissions. Additionally, the classification of iOS apps as malicious or benign is performed using frequent permissions. Our experimental results show that frequent permission-based classification gives better precision in comparison to the classification performed by total permissions.