A New Approach for Early Detection of Vulnerabilities Across the Automotive Software Development

This paper presents a new approach of evaluating automotive development processes, starting from the early stages of development to the final product. The main objective of this paper is to ensure transparency among stakeholders and provide accurate feedback throughout the development cycle, with the ultimate goal of preventing software defects and conflicts that can be reproduced. The concept aims to enable stakeholders to efficiently evaluate the status and quality of software releases. The latter can be achieved by an intelligent recommendation system that selects and poses a few precise questions from a large database during each development cycle, minimizing time demands on development stakeholders while obtaining important results. To accomplish this, we have extracted the most crucial development metrics from the literature and transformed them into questions. To each question is assigned a weight based on its importance, and each answer is given a corresponding value. Through this process, we have created a dynamic pool of potential questions that can be selected using the reinforcement learning algorithm of Contextual Multi-Armed Bandits (CMAB), which serves as a recommendation system to seven key roles among development stakeholders. The software implementation of this method enables not only the escalation of the number of components under evaluation but also the continuous improvement of its performance over time.