Security mobile app user reviews

Elaborating security requirements (SRs) at the early stage of requirements engineering is essential in software engineering. It helps develop a secure system and takes security into account in all stages of the software development life cycle. Most of the SRs are written based on security engineering and developers' perspectives. Today, mobile app user reviews have valuable information related to all the issues facing end users while using an application. Most of this information is related to requirements engineering that covers functional and non-functional requirements. Non-functional requirements can be elaborated on many properties of an application (e.g., security, reliability, and safety). Performing a manual review of mobile app user reviews to extract reviews related to security issues that can be used to express important SRs is time-consuming and susceptible to human errors. In this paper, we develop a transformer-based model to identify mobile app user reviews related to SRs. We collect more than one Million mobile app user reviews. Then, we annotate them utilizing an active learning approach. Additionally, we utilize several AI techniques such as natural language processing and sentence transformers. The results show that the developed model can predict security-related mobile app user reviews with an average accuracy of 87.5\%.