GENDroid: A Optimized Hybrid Android Malware Detection Framework via Multimodal Feature Fusion and Genetic Algorithm

The rapid growth of Android applications has been accompanied by a surge in malicious apps that exploit system permissions and static components to bypass conventional security defenses. While static analysis has proven useful for early malware detection, many existing approaches fall short due to limited feature representation and a lack of adaptability to evolving threats. This paper introduces GENDroid, a novel framework that enhances Android malware detection by combining multi-modal static feature fusion with evolutionary optimization. GENDroid integrates diverse features—including permissions, API calls, and intent filters—into a unified representation, enabling a deeper understanding of application behavior. To optimize both feature selection and classifier performance, a Genetic Algorithm (GA)-driven strategy is employed, allowing the system to evolve and adapt automatically. The optimized feature set is used to train an ensemble of machine learning classifiers on a comprehensive dataset comprising both real-world and synthetic Android applications. Experimental results demonstrate that GENDroid delivers high detection accuracy, significantly reduces false positives, and remains robust against adversarial variants. Its modular design also allows for the seamless integration of additional static or behavioral features. By intelligently combining feature diversity with adaptive learning, GENDroid provides a practical and scalable solution for Android malware detection, effectively addressing the limitations of traditional static analysis techniques.