STATIC ANALYSIS TECHNIQUES FOR ANDROID BASED MALWARES

Abstract

The number of threats has also gone up a lot as Android apps have become more popular. This puts users at risk of a lot of different kinds of malware that can steal data, use resources and invade privacy. This thesis shows how to use static analysis to find bad Android apps by creating a framework for finding threats. The method works by getting static information from application packages (APKs), like declared permissions, suspicious API calls and hardcoded strings (like IP addresses and URLs). This lets you look for possible threats without having to run the program. We looked at a dataset that had both good and bad examples of malware. We used the Synthetic Minority Over-sampling Technique (SMOTE) to fix the class imbalance. We showed a number of deep learning and machine learning classifiers how to work with feature vectors. The Support Vector Machines, Decision Trees, Random Forests, Logistic Regression, K-Nearest Neighbors, Gradient Boosting, AdaBoost, XGBoost and Multi-Layer Perceptron (MLP) were all used. We used common metrics like accuracy, precision, recall and F1-score to test these models. The experimental results show that the ensemble-based methods and the MLP classifier were the best at telling the difference between good and bad apps, with an accuracy rate of over 98%. The study shows that strong algorithms and static threat detection methods can work together to arrange malware in a very useful and efficient way. The suggested framework is a quick and easy way to find Android threats. It will also be possible to use hybrid or dynamic analysis methods in the future.