Zero-Vuln: Using deep learning and zero-shot learning techniques to detect zero-day Android malware

Abstract

The prevalence of cyber security threats, such as the Android Zero-day vulnerability, is becoming increasingly worrisome. With the widespread use of Android-powered mobile devices, attackers are leveraging zero-day vulnerabilities to infect Android software at an alarming rate. Detecting zero-day vulnerabilities in Android applications is particularly challenging due to their unpatched and undiscovered nature, resulting in a lack of reference points for identification. In response to this issue, we propose a novel and effective system called Zero-Vuln, which is designed to classify and identify zero-day Android malware. Zero-Vuln leverages deep learning and zero-shot learning techniques, as well as established data-sets, to identify previously unknown malware. Our approach achieves a remarkable performance of 83% accuracy, as well as high precision and recall, and represents a significant contribution to the field of cyber security.