SIGPID For Machine Learning based Android Malware Detection

Abstract

Smart phone use has been gradually growing in recent years, as has the number of Android device users. As the number of Android app users grows, malicious Android apps are being developed as a tool to steal sensitive data and commit identity theft / fraud on mobile banks and wallets. There are a plethora of malware identification tools and apps on the market. However, new complex malicious apps generated by intruders or hackers need powerful and efficient malicious application detection tools. To begin, we must collect a dataset of prior malicious apps as a training set, and then compare the training dataset to the trained dataset using the CNN algorithm and the RNN algorithm. Open source datasets, such as Kaggle datasets, were used to build the datasets. We use a pre-processing and attribute extraction technique before running the algorithm. Preprocessing of data that is related to independent variables or data features. It ultimately assists in the normalisation of data within a specified boundary. Standard scalar data is usually distributed within each function, and will scale them to the point where the distribution is zero and the root mean square deviation is one, feature extraction techniques such as the tf-idf transform and data pruning are used. It also aids in the acceleration of algorithmic calculations. Using this algorithm, we can detect threatful Mobile applications.