Pattern lock screen detection method based on lightweight deep feature extraction

In the digital age, many people have used mobile phones, thus, mobile phones are one of the most commonly used crime tools. Users can take security measures to their mobile devices using various authorization methods such as passwords or screen patterns (touch screen security authentication). The used security measures generally make mobile forensic analysis difficult or even impossible. In order to overcome this problem, a novel intelligent pattern lock detector is presented in this research. The proposed lock detector uses transfer learning to extract deep lightweight features, iterative feature chosen function and a shallow classifier. A feature extraction network has been created by using SqueezeNet and MobileNet-V2, which are among the deep learning architectures in this work. Iterative Minimum Redundancy Maximum Relevance (ImRMR) was used for feature selection. Linear discriminant analysis (LDA) was selected for the classifier. The proposed model has been developed on three image datasets. These datasets are named clean, slightly dirty and medium dirty. 99.75%, 98.55% and 96.50% classification accuracies have been reached on the used three datasets, respectively. The findings clearly denote that the success of the presented deep lightweight features and ImRMR-based detector.