PFI-TT: An Analysis Tool Supporting the Safe Deployment of New Features in Evolving Software Systems

该创新技术转化伙伴关系（PFI-TT）项目的广泛影响/商业潜力将开发一个工具支持的环境，旨在帮助软件密集型系统的安全分析。为医疗输液泵、阳性列车控制和机器人应用开发高度可靠的软件是一个具有挑战性的过程，需要严格和系统的危害分析。这样的系统传统上是在一个精心控制的过程中开发的，强调详细的计划、前期设计和基于阶段的质量保证网关。尽管做出了这些努力，产品还是会因为危及生命的安全漏洞而被召回。开发安全系统的挑战导致了所谓的“大冻结”现象，即引入新功能的成本、努力和风险抑制了产品的发展。建议的解决方案将帮助软件安全分析师和其他涉众识别系统级更改，同时为认证或内部审计目的构建安全案例。这些能力使组织能够在不牺牲安全、保障或其他关键问题的情况下发展他们的产品，从而增加他们的竞争优势。该项目培训妇女、少数民族和残疾人等各类骨干从事创业活动。这个项目将利用尖端的深度学习技术来训练一个领域特定的语言模型，该模型将用于自动生成和发展工件（如需求、设计、代码和测试用例）之间的精确跟踪链接。生成的跟踪链接将用于生成新的危害及其缓解途径的可视化（即危害切片），并突出显示跨系统不同版本引入的更改。此外，风险切片将集成到现有的安全分析技术中，如故障树和安全保证案例。新的基于危险的可视化设计是为了帮助分析人员和开发人员在大型、复杂和不断发展的软件项目中处理紧急的安全问题。最后，这项工作将提供工具来分析软件系统中的变化，生成解释以帮助识别潜在的安全问题，并建议可以采取的行动来减轻变化带来的安全风险。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

The broader impact/commercial potential of this Partnerships for Innovation - Technology Translation (PFI-TT) project will develop a tool-supported environment designed to aid in the safety analysis of software intensive systems. Developing highly dependable software for medical infusion pumps, positive train controls, and robotics applications is a challenging process that requires rigorous and systematic hazard analysis. Such systems are traditionally developed within a carefully controlled process that emphasizes detailed planning, upfront design, and phase-based quality assurance gateways. Despite these efforts, products are recalled for life-threatening safety vulnerabilities. The challenges of developing safety systems have led to the phenomenon known as the "big freeze" in which the cost, effort, and risk of introducing new features inhibit product evolution. The proposed solution will aid software safety analysts and other stakeholders in identifying system-level changes while simultaneously building a safety-case for certification or internal audit purposes. These abilities enable an organization to evolve their products without sacrificing safety, security, or other critical concerns, thereby increasing their competitive advantage. The project trains a diverse cadre of women, minorities, and persons with disabilities in entrepreneurial activities. This project will utilize cutting-edge, deep learning technologies to train a domain-specific language model that will be used to automatically generate and evolve accurate trace links between artifacts such as requirements, design, code, and test cases. The generated trace links will be used to generate novel visualizations of hazards and their mitigation pathways (i.e., hazard slices), and to highlight changes that have been introduced across different versions of the system. In addition, the hazard-slices will be integrated into existing safety-analysis techniques such as fault-trees and safety-assurance cases. The novel hazard-based visualizations are designed to aid analysts and developers in addressing emergent safety concerns in large, complex, and evolving software projects. Finally, the work will provide tools for analyzing changes in the software system, generating explanations to aid in the identification of potential safety problems, and recommending actions that could be taken to mitigate safety risks introduced by the changes.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.