INPERMHARDROID: ANDROID MALWARE DETECTION USING INTENTS, PERMISSIONS AND HARDWARE COMPONENTS

Abstract

As mobile devices continue to evolve and become integral to daily life, the risks posed by malicious software are increasing. Cybercriminals continuously develop more sophisticated malware, bypassing conventional security mechanisms and compromising user privacy. To counter these threats, it is crucial to establish a highly efficient and precise malware detection framework. This study introduces a creative approach for malware identification and classification. The framework employs Fisher’s P-test, a statistical method that isolates the most distinguishing features between benign and malicious applications, eliminating irrelevant attributes early in the process. This initial refinement improves efficiency for subsequent classification. Following this, the Relief algorithm further enhances the data set by identifying the most relevant attributes that contribute significantly to malware detection. The integration of these two techniques results in an optimized dataset, reducing complexity while improving classification accuracy. Once the features are refined, machine learning models analyze the data to detect and categorize malware. Among the classifiers evaluated, Random Forest proves to be the most effective, utilizing an ensemble of decision trees to enhance predictive accuracy while minimizing overfitting. When applied to a dataset combining permissions, intents, and hardware components, the Random Forest classifier achieves an impressive precision of approximately 97%, demonstrating the effectiveness of a multi-dimensional approach in malware detection. This research makes substantial contributions to the field of mobile security by designing a more precise and interpretable malware detection system. As threats targeting mobile platforms continue to grow, integrating advanced security techniques is essential for protecting user data and privacy. This study provides valuable insights that pave the way for advancing machine learning-driven cybersecurity strategies, strengthening mobile ecosystems against emerging cyber threats. Index Terms—Mobile Malware, Android Security, Fisher’s P test, Relief Algorithm, Feature Selection.