ANDROID MALWARE DETECTION USING GUMBEL-ATTENTION FEATURE SELECTOR NETWORK

Abstract

With Android running on billions of devices globally, it has emerged as the founda tion of the mobile industry. However, it has also become a prime target for malware attacks beacause of its open-source nature and diverse ecosystem. Android’s permissions, intent mechanisms, and hybrid components are frequently used by malicious apps to obtain sen sitive data without authorization or alter device functionality. As attackers employ more obfuscation techniques and adversarial strategies to avoid detection, existing malware de tection methods—such as static and dynamic analysis—find it difficult to keep up. In order to overcome these obstacles, we present a brand-new framework called GAFS-Net (Gumbel-Attention Feature Selector Network), which uses sophisticated feature selection and attention mechanisms to improve Android malware detection. In order to find the most important features while eliminating irrelevant data, GAFS Net cleverly analyzes big datasets.It uses Gumbel-Softmax-based selection to rank hybrid components, permissions, and intents based on how relevant they are to identifying malicious activity. In order to improve classification accuracy and inter pretability, the framework also incorporates at tention mechanisms, which guarantee that the most significant features are given priority. Our tests show that GAFS-Net performs well on three datasets: intents, permis sions, and hybrid components, with an astounding 96% accuracy rate. GAFS-Net simplifies the detection process and produces more dependable results than conven tional techniques, which frequently struggle with noisy data and ineffective feature prioritization. Further more, because of its transparency, security researchers are better able to comprehend how malware functions, which aids in the development of preventative cybersecurity measures. GAFS-Net offers a workable solution for malware detection in the real world be cause of its high performance, scalability, and modular design. As Android threats continue to evolve, frameworks like GAFS-Net open the door for more sophisticated security systems, guaranteeing improved protection for users and their data.