CSR: Small: Collaborative Research: Safety Guard: A Formal Approach to Safety Enforcement in Embedded Control Systems

CSR：小型：协作研究：安全卫士：嵌入式控制系统中安全执行的正式方法

• 批准号: 1813117
• 负责人: Chao Wang
• 金额: $ 25万
• 依托单位: University of Southern California
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2018
• 资助国家: 美国
• 起止时间: 2018-08-01 至 2022-07-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1813117&HistoricalAwards=false
• 关键词: CSR; Small; Collaborative; Research; Safety

Ensuring the safety of embedded control systems, such as the ones used in cars, is remarkably difficult as shown by the many recalled cars in recent years, causing tens of billions of financial loss each year. This project aims to improve the safety of critical components in embedded control systems by developing safety guards, a reactive component generated automatically from safety requirements and attached to the original system, to ensure the combined system is safe even if the original system violates the safety requirements. The intellectual merit of this project lies in the set of methods and tools to be developed for synthesizing safety guards of black-box systems. Specifically, this project consists of three research tasks: (a) building a benchmark suite of critical components and their safety requirements; (b) developing synthesis algorithms for constructing the finite-state machines (FSMs) of the safety guards; and (c) developing software synthesis tools for automatically generating software code that implements these FSMs. This research project will benefit a wide range of application domains, including automotive and avionics, which will be investigated through collaborations with industry. It will help improve the safety of critical components, including those based on machine learning and artificial intelligence techniques. It will simplify certification since the relatively simple safety guard can be certified against safety requirements in place of the detailed model of a critical component. And last but not the least, it will simplify the development process by allowing people to focus on functionality and performance without worrying about safety violations at the same time. The resulting software tools, together with evaluation benchmarks and experimental data, will be made available to the public. To facilitate dissemination and sharing, the project will maintain online documentations, tutorials, slides, and source code of the tool and benchmark repositories. Besides the research websites of participants at Virginia Tech and University of Southern California, the following website will be dedicated to disseminate the project results broadly: http://chaowang-vt.github.io/safetyguard/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.

确保嵌入式控制系统的安全，比如汽车上使用的控制系统，是非常困难的，正如近年来许多召回汽车所表明的那样，每年造成数百亿美元的经济损失。该项目旨在通过开发安全卫士来提高嵌入式控制系统中关键组件的安全性，安全卫士是根据安全需求自动生成的并附加到原始系统的反应性组件，以确保组合系统即使在原始系统违反安全要求的情况下也是安全的。这个项目的智力价值在于为合成黑匣子系统的安全防护而开发的一套方法和工具。具体地说，该项目包括三项研究任务：(A)建立关键部件及其安全要求的基准套件；(B)开发用于构建安全卫士的有限状态机(FSM)的综合算法；以及(C)开发用于自动生成实现这些FSM的软件代码的软件综合工具。这一研究项目将使包括汽车和航空电子在内的广泛应用领域受益，这些领域将通过与工业界的合作进行研究。它将有助于提高关键部件的安全性，包括基于机器学习和人工智能技术的部件。它将简化认证，因为相对简单的安全卫士可以根据安全要求进行认证，而不是关键部件的详细模型。最后但并非最不重要的一点是，它将简化开发过程，使人们能够专注于功能和性能，而无需同时担心违反安全规定。由此产生的软件工具以及评价基准和实验数据将向公众提供。为便利传播和分享，该项目将维护在线文件、教程、幻灯片以及该工具和基准储存库的源代码。除了弗吉尼亚理工大学和南加州大学参与者的研究网站外，以下网站将致力于广泛传播项目结果：http://chaowang-vt.github.io/safetyguard/This奖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智力优势和更广泛的影响审查标准进行评估，被认为值得支持。

项目成果

Synthesizing Fair Decision Trees via Iterative Constraint Solving

通过迭代约束求解综合公平决策树

• 发表时间: 2022
• 期刊: International Conference on Computer Aided Verification
• 作者: Wang, Jingbo;Li, Yannan;Wang, Chao
• 通讯作者: Wang, Chao

DiffRNN: Differential Verification of Recurrent Neural Networks

• DOI: 10.1007/978-3-030-85037-1_8
• 发表时间: 2020-07
• 作者: Sara Mohammadinejad;Brandon Paulsen;Chao Wang;Jyotirmoy V. Deshmukh

LinSyn: Synthesizing Tight Linear Bounds for Arbitrary Neural Network Activation Functions

• DOI: 10.1007/978-3-030-99524-9_19
• 发表时间: 2022-01
• 期刊: ArXiv
• 作者: Brandon Paulsen;Chao Wang

Example Guided Synthesis of Linear Approximations for Neural Network Verification

• DOI: 10.1007/978-3-031-13185-1_8
• 发表时间: 2022
• 作者: Brandon Paulsen;Chao Wang

Shield Synthesis for Real: Enforcing Safety in Cyber-Physical Systems

真实的护盾合成：增强网络物理系统的安全性

• DOI: 10.23919/fmcad.2019.8894264
• 发表时间: 2019
• 期刊: 2019 Formal Methods in Computer Aided Design (FMCAD
• 作者: Wu, Meng;Wang, Jingbo;Deshmukh, Jyotirmoy;Wang, Chao
• 通讯作者: Wang, Chao