SaTC: EDU: Secure and Private Artificial Intelligence

随着人工智能（AI）被整合到更多的系统中，与隐私保护相关的网络安全问题越来越多，同时也需要训练有素的专业人员来开发和部署值得信赖的人工智能系统。该项目的目标是开发教学材料，通过现实世界的例子，让学生了解人工智能系统固有的隐私问题。该项目将通过解决日益增长的国家对训练有素的专业人员的需求，从而产生直接和长期的影响，这些专业人员能够采取整体方法解决复杂人工智能系统中现实世界的问题。此外，该项目将使来自代表性不足的少数群体的众多学生受益，并改善劳动力的多样性。佐治亚州立大学是一所为少数族裔服务的机构，与亚特兰大大都会和更广泛地区的传统黑人学院和大学（HBCUs）有着密切的联系。该项目将通过为教育界举办的培训讲习班传播已开发的材料，从而促进采用涉及隐私和人工智能的培训模块。跨学科项目团队将开发一门名为“私人人工智能”的新课程，该课程将包括教学模块和实践实验室，采用最先进的私人人工智能技术来解决人工智能系统的不同隐私挑战。这些教学模块将根据学习科学原则，特别是以问题为中心的教学（PCI）原则来设计。这些模块将基于现实世界的系统，旨在系统地涵盖人工智能系统的基本隐私原则和实用技能。互动课程活动的学习目标是让学生(i)了解私人人工智能的基本概念和原理；（ii）了解私隐攻击和防御、不同的私隐保护技术，以及每种方法的优缺点；以及（iii）获得开发和部署私人人工智能系统的知识和技能。可交付成果还包括帮助教师将模块集成到课程中的手册，以及在课堂上实施PCI活动的指南，因为许多教师不是教学科学方面的专家。为了简化集成并鼓励采用，模块和实验室将基于开源软件和工具，这些软件和工具可免费用于教育目的。这些模块也将通过免费的云平台分发。该项目由安全与可信网络空间（SaTC）计划支持，该计划资助解决网络安全和隐私问题的提案，在这种情况下，特别是网络安全教育。SaTC项目与《联邦网络安全研究与发展战略计划》和《国家隐私研究战略》保持一致，旨在保护和维护网络系统日益增长的社会和经济效益，同时确保安全和隐私。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。

As artificial intelligence (AI) is incorporated into more systems, there are growing cybersecurity concerns related to privacy protection as well as a need for highly trained professionals who can develop and deploy trustworthy AI systems. The goal of this project is to develop instructional materials that expose students to privacy issues inherent in AI systems through real-world examples. The project will have a direct and long-term impact by addressing the growing national need for highly trained professionals capable of taking a holistic approach to solving real world problems in complex AI systems. In addition, the project will benefit numerous students from underrepresented minority groups and improve diversity of the workforce. Georgia State University is a minority-serving institution and has strong connections with Historically Black Colleges and Universities (HBCUs) in Metro Atlanta and the wider region. The project will disseminate the developed materials through training workshops for the educational community and therefore promote adoption of training modules addressing privacy and AI.The interdisciplinary project team will develop a new course, “Private AI”, which will include instructional modules and hands-on labs that employ state-of-the-art private AI techniques addressing different privacy challenges of AI systems. These instructional modules will be designed according to learning science principles, specifically the principles of problem-centered instruction (PCI). The modules will be based on real-world systems and are designed to cover fundamental privacy principles in AI systems and practical skills systematically. The learning objectives of the interactive curricular activities are for students to (i) understand fundamental concepts and principles of private AI; (ii) understand privacy attacks and defenses, different privacy-preserving techniques, and the pros and cons of each approach; and (iii) gain knowledge and skills in the development and deployment of private AI systems. The deliverables also include manuals to help instructors integrate the modules into their curricula and guidelines on implementing PCI activities in the classroom since many instructors are not experts in instructional sciences. In order to simplify integration and encourage adoption, the modules and the labs will be based on open-source software and tools that are free to use for educational purposes. The modules will also be distributed via free cloud platforms. This project is supported by the Secure and Trustworthy Cyberspace (SaTC) program, which funds proposals that address cybersecurity and privacy, and in this case specifically cybersecurity education. The SaTC program aligns with the Federal Cybersecurity Research and Development Strategic Plan and the National Privacy Research Strategy to protect and preserve the growing social and economic benefits of cyber systems while ensuring security and privacy.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.