Chat-GPT for Android malware detection

Abstract

The use of large-language models (LLMs) in the field of cybersecurity has been increasing greatly in recent years. With the advent of ChatGPT by OpenAI, there have been many different use cases for LLMs in cybersecurity, including in intrusion detection, as well as in vulnerability detection. However, there has yet to be much research done in the use of LLMs for malware detection, more specifically, in the area of Android malware detection. In this paper, we will look at how we can capitalise on the use of ChatGPT in detecting malware or malicious source code in Android applications. We will devise various prompts and include a framework design that will allow ChatGPT to detect Android malware code. We will also propose a hierarchical structure to evaluate the effectiveness of ChatGPT in Android malware detection. This hierarchical structure aims to understand the important pieces of information which are present in malware applications, that are needed by ChatGPT to detect malicious pieces of code in Android applications. In the study, we found that the manifest files are sufficient for ChatGPT to detect malicious code in 68% of a specific malware family. Through this study, we will be able to understand how ChatGPT is able to detect malware and understand the reasons for failing to detect.