Detection of Malware in Android Phones Using Machine Learning
K J Poojitha · International Journal for Research in Applied Science and Engineering Technology · 2022
Abstract: In a major cyber security scare, around 1.5 crore Android devices in India have been infected by malware without the knowledge of the users. According to a report by cyber security solution firm Check Point Research in 2020, a new variant of mobile malware has quietly infected around 2.5 crore devices worldwide. Malware is any type of malicious software or code designed to harm a user's device, such as trojans, adware, ransomware, spyware, viruses or phishing apps. The permissions and API-calls are extracted from all Android applications, and both were included as features in the dataset. It is a process of analysing the malware binary without running the code. To offer a simple, streamlined, document-centric experience Jupyter Notebook interactive development environment and Flask is utilized. The Androguard tool and genetic algorithm analyses the APK files by separately extracting the permissions for each APK file. Supervised Machine Learning algorithms used are Support Vector Machine (SVM) and MLP an ANN numeral network approach is used to compare the traditional machine learning techniques. Experiments will be conducted on two types of models, traditional machine learning classifiers and deep learning neural networks. Initially, the classifiers are trained using the dataset, taken from android malware dataset and then testing and evaluationis performed based on the extracted features. An efficient method to detect the presence of malware in the android mobile phones using the permissions, API calls is implemented and the best classifier is identified which gives optimal results with accuracy, F-measure, Recall, and Precision scores. This would enable users to easily navigate various resources available with an adaptive user interface using android application.